• H2 laser arrived, unloaded in large item access area
• Reiboldt construction at EX
• Usual chamber cleaning/SUS/SEI test stand work

Split crew: Carolyn, Chris, and Scott H. played support roles today.

Mark Layne and I removed bellows protection and the started the fiber hunt on the support tubes. I took a few pix: please see below. We collected several of the more flamboyant fibers just in case anyone wants to look at them later. 

Then, it was time for vacuuming: bellows, ring, lower chamber, and chamber floor.

Wipe down was started after lunch. Mark and I agreed on locations for Rai's wipe samples, so we will collect those as we go. Wipe down of the collar was completed.

Some tooling and staging work was also completed today. Zack finished taking apart all the drills that seized when we tried to use them in this chamber. The parts have been sorted and will be cleaned. Then, we will try the LLO random re-assembly method on those. The drills that worked in chamber have been set aside for now. Randy and Mick moved the pedestals for supporting the dome and staging tables down to Y-end.

  We took a decent set of system id over the weekend and it looks like all of those people who were snickering at Brian are going to have to aplogize. The signal from rX andrY stage one actuators to rX and rY stage 2 CPS is about 50-100 times larger then the signal for the other four degrees of freedom, and since we have not normalized the actuators yet we expect a few percent of cross coupling at low frequency. The results for Stage 2 Act to stage 1 CPS are more or less the same

J. Kissel, K. Arai, R. Lane, J. Garcia, R. Mittleman, B. Shapiro

In eLIGO, the SUS watchdog had been cobbled together as its own CDS block quickly using iLIGO watchdogs as a model, and was never properly understood (at least by me), nor well documented. In the switch over to aLIGO RCG, I had made an attempt to simplify the SUS watchdog by taking advantage of the more simple CDS RMS filter block (see entry 1265 for details).

However, I had naively used the block to take the RMS of the OSEM raw input signals and called it good, because I had assumed that the time constant of the averaging filter also defined the frequency response of what signal was being watched (as though the RMS filter was a linear block). However, after some experience using this simple watchdog, and discussions with Brett, Rich, and Koji, they've [reminded / convinced / taught] me that this isn't what we want to watch at all.

Hence, we've installed band-limiting filters between the raw OSEM signals and the RMS filter. The frequency response of this filter is attached, but can be described by the following bit of design code:

bandLimPoles = [0.1 10 10];
bandLimZeros = [0];
bandLimFilter.c = zpk(bandLimZeros,-2*pi*bandLimPoles,1);
gain = abs(squeeze(freqresp(bandLimFilter.c,2*pi*1)));
bandLimFilter.c = bandLimFilter.c / gain;

The motivation behind this shaping of the signal is as follows:
- The frequency band in which the suspensions are moving the most is between 0.1 Hz and 10 Hz, given that all resonances for all degrees of freedom are confined to this band.
- Above 10 Hz, the OSEM signal is dominated by sensor noise, and hence is not useful as a watchdog signal. 
- Even if there were useful signals above 10 Hz, (a) the coils and coil drivers do not have the strength to push the suspension around, (b) the suspension's response to actuation is rolling off drastically anyways, and (c) given that this frequency band is above the fundamental resonances of the suspension, if the suspension point is driven at 10Hz and above, the significant passive isolation should provide sufficient protection.
- The OSEM signal must be AC coupled, or else the RMS is dominated by the DC offset of the OSEM which is not useful as a watchdog signal. We are interested in the position amplitude RMS around the resonances, such that we can catch large, on-resonance, amplitude motion that would drive the suspension into its stops, In other words the response of the signal fed into the RMS filter should be flat, since OSEMs are inherently a position sensor. That means there must be *some* pole frequency where we switch from an AC coupled velocity RMS to flat position RMS. If this pole frequency is too low, the position RMS of the resonaces will be washed out by the inherently large low-frequency motion. Hence, as a first, reasonable guess, we've chosen 0.1 Hz -- just below the lowest longitudinal resonance.
- As mentioned in aLOG 1265, the step response of the RMS filter is roughly 7 seconds. This is another goldy-locks problem, in that if the step response is too long or too short then one loses functionality. We know 100 seconds is too long, 1 second is too short. Given that we're "stuck" with 7 seconds we figured it would suffice for now.

In summary, the exact shape (where to put the poles and zeros) of the signal to be RMS'd is quantitatively arbitrary, but the response now has the right qualitative shape. Poles at 0.1 and 10 Hz can be tweeked later if necessary, and given increased functionality in the RMS block we could adjust the time constant. The watchdog now has the right functionality for the system at hand .

I also attach pictures of the infrastructure
- osemwatchdogsimulink.png: screenshot of the signal flow of an OSEM sensor flag generator in the QUAD_MASTER.mdl simulink diagram (found in the QUAD//WD/OSEM/ subsystem, isoStage=[M0, R0, L1, L2]) 
- osemwatchdogmedm.png: screen shot of a given stage's watchdog MEDM screen. The OSEM BAND LIMITERS bank is button that takes you to
- osembandlimmedm.png: the filter banks where the band limiters are defined in FM1.

The last attachment is a spectra of an L1 OSEM comparing the signal before and after the bandlimiting filter, just to prove it's functionality. The suspension is locked and not plugged in, so the actual signal is rather meaningless. Once we get a functional suspension, I'll put the effort into making a calibrated plot, containing more useful information,

J. Kissel, R. Lane, K. Arai, J. Garcia, R. Mittleman

An integral part of the suspension watchdog system is the aLIGO RCG's cdsRms block (design details to follow), so I wanted to give it a little extra attention. As of RCG 2.3 and later, this block is implemented with the following C Code:

1  float rms;
2  static float rms_avg;      // which means rms_avg is a persisent variable between loop iterations
3  if(feInit)                 // If it's the first loop cycle
4  {
5       rms_avg = 0.0;
6       } 
7  else 
8  {
9       // RMS: RMS
10       rms = ;
11       if(rms > 200000) rms = 200000;
12       if(rms < -200000) rms = -200000;
13       rms = rms * rms;                            // HERE'S THE SQUARE
14       rms_avg = rms * .00005 + rms_avg * 0.99995; // HERE'S THE LOW PASS, "MEAN"
15       rms = lsqrt(rms_avg);                       // HERE'S THE SQUARE ROOT
16        = rms;
17       }



Of particular interest here is the low pass filter that defines the averaging time constant (defined on line 14). Given the above c-code, the filter is defined in a standard format, with the following loop structure (taking x = rms * rms, A = 0.00005, and y = rms_avg):
       +---+
 x ----| A |-----( + )----------------------------- > y
       +---+       |                         |
                   |                         |
                   |  +-------+  +--------+  |
                   +--| (1-A) |--| z^(-1) |--+
                      +-------+  +--------+
which should nominally have a transfer function H(f) of 

y = A * x + y * (1-A) * exp(- 2 * pi * i * f * T)      // identical to line 14!!
y * (1 - (1-A) * exp(- 2 * pi * i * f * T) ) = A * x
H(f) = y / x = A / (1 - (1-A) * exp(- 2 * pi * i * f * T) )

where f is the frequency vector and T = 1/fs is the sampling time. This transfer function corresponds to a single pole at 

f = 1/(2*pi*T)*log(1/(1-A))

This means for a fixed A, the averaging pole is dependent on the sampling rate fs of the model using the RMS block. For future upgrades of the RCG, we should consider having an additional input to this block that defines A, such that the user can customize the filter time constant where needed, and have it adaptable to other model sampling rates.

For the suspension models, running at fs = 16384 Hz, this corresponds to a pole frequency of 0.13 Hz, indicating that the RMS block has a impulse response of about 7 sec. This is coincidentally just about the right response for a suspension watchdog, at least as a first iteration. (Further design details to come).

K. Arai

Based on a discussion with Jeff K, change of the channel names from H2:SUS-*-*-OSEMOUTF-* to H2:SUS-*-*-COILOUTF-* was carried out.
The point is that this naming should be better as they are actually connected to the coils, not to the OSEM sensors.

- ITMX/ITMY/FMX/FMY/BS models were influenced and action has been made.
- Scripts using OSEMOUTF are to be influenced. (need attention)
- Related simulink models were updated and running with green status
- Related MEDM screen were updated
- Related filter definitions (.txt files), daq configuration for slow and fast channels (.ini files) were updated. (i.e. the same filters as before have been loaded).

Detailed procedure

1. Update the simulink models for QUAD and BSFM

Any "OSEMOUTF" related names and descriptions were changed to "COILOUTF" and so on.
The modification has been done on the following files in /opt/rtcds/lho/h2/userapps/release/sus/common/models/

BSFM_MASTR.mdl QUAD_MASTR.mdl

2. Replace the filter names in the "chans" file

Before building and installing the new models, the filter names in the "chans" file should be replaced
as the installation script tries to inheret the previous configuration. If this replacement is not done before the installation,
the previous filter coefficients are lost in the new files, and the entry for the new blank filters are created.
Once this happens, we will have to move the filter definitions from the old files with foton (one by one).
This is painful. So we need this operation!

Any instances of "OSEMOUTF" in the following files are replaced to "COILOUTF"
(in /opt/rtcds/lho/h2/chans/)

H2SUS_ITMX.txt H2SUS_ITMY.txt H2SUS_FMX.txt H2SUS_FMY.txt H2SUS_BS.txt

3. Replace the channel names in the daq configuration files

Also the DAQ configuration should be taken over at the build for the same reason.
This should be done for both of the fast and slow channels.

Any instances of "OSEMOUTF" in the following files are replaced to "COILOUTF"
(in /opt/rtcds/lho/h2/chans/daq/)

Slow channels:
H2EDCU_SUSITMX.ini H2EDCU_SUSITMY.ini H2EDCU_SUSFMX.ini H2EDCU_SUSFMY.ini H2EDCU_SUSBS.ini

Fast channels:
H2SUSITMX.ini H2SUSITMY.ini H2SUSFMX.ini H2SUSFMY.ini H2SUSBS.ini

4. Bulld the models

h2susitmx, h2susitmy, h2susfmx, h2susfmy were built and installed. I have noticed that there has been no BS model yet.

5. Restart the realtime codes

At h2susb478, h2iopsusb478, h2susitmx, h2susitmy have been restarted.
At h2susb78, h2iopsusb78, h2susfmx, h2susfmy have been restarted,

6. Restart the data concentrator

The daq status signals returned to green.

7. Update MEDM screens

The medm screens in the following directories were to be updated
/opt/rtcds/lho/h2/userapps/release/sus/common/medm/QUAD/
/opt/rtcds/lho/h2/userapps/release/sus/common/medm/BSFM/

  a. The files which contains OSEMOUT in their names were renamed

(in BSFM)
SUS_CUST_BSFM_M1_COILOUTF.adl SUS_CUST_BSFM_M2_COILOUTF.adl
(in QUAD)
SUS_CUST_QUAD_L1_COILOUTF.adl SUS_CUST_QUAD_L2_COILOUTF.adl SUS_CUST_QUAD_M0_COILOUTF.adl SUS_CUST_QUAD_R0_COILOUTF.adl

  b. change any instances of "OSEMOUT" to "COILOUT".
  This includes the replacement of the channel names as well as the filenames.

(in BSFM)
SUS_CUST_BSFM_OVERVIEW.adl SUS_CUST_BSFM_M1_COILOUTF.adl SUS_CUST_BSFM_M2_COILOUTF.adl
(in QUAD)
SUS_CUST_QUAD_OVERVIEW.adl SUS_CUST_QUAD_R0.adl SUS_CUST_QUAD_L1_COILOUTF.adl SUS_CUST_QUAD_L2_COILOUTF.adl SUS_CUST_QUAD_M0_COILOUTF.adl SUS_CUST_QUAD_R0_COILOUTF.adl

8. Confirm dataviewer and each medm screen shows right things

9. Commit the Simulink models and MEDM screens to the svn

Summary. Field trials are about to begin for elimination of Time Error Correction (TEC), the practice of running the power grids at 59.98 or 60.02 Hz for short periods to keep 60 Hz clocks accurate. This change in 60 Hz regulation should be neutral or help with detection of the Crab pulsar at 59.4 Hz because it is expected to reduce low frequency excursions and the new average grid frequencies are expected to be higher, about 60.002 (LLO) and 60.0009 (LHO).


The North American Electric Reliability Corporation (NERC), is planning on changing the regulation of the 60 Hz frequency for the eastern (LLO) and western (LHO) power grids, raising questions about how this may effect detection of the Crab by changing the tails of the 60 Hz peak.  I talked with Andy Rodriquez (andy.rodriquez@nerc.net), Director of Standards Development at NERC, and reviewed NERC documents to asses the proposed elimination of Time Error Correction, which would mean that, over the long term, the average line frequency would not be precisely 60 Hz.

The power grid frequency varies from 60 Hz when supply and demand become unbalanced. When power use increases with morning activities, or a generator shuts down, the frequency decreases. Figure 1 shows a low frequency excursion from loss of generation. As demand drops off at night or as more generators come on line, the frequency increases. Figure 2 shows that morning and night are particularly susceptible to frequency excursions. Figure 3 shows the rms deviation from 60 Hz for the different grids in 1996.

Several control mechanisms affect the line frequency and are used either to minimize frequency departures from 60 Hz or to adjust financial balances between the generating authorities. Organized by time scales, the control mechanisms are:  1) primary control, on a second to minute scale, consisting of independent generator governors that increase or decrease, e.g., the flow of dam water to turbines when a deadband of, typically, 36 mHz is exceeded. Primary control also includes Automatic Generation Control (AGC), automatic control of clusters of generators,  2) secondary control, on a scale of 1 – 10 minutes, includes AGC and manual phone calls to power plants and other load management actions, 3) tertiary control, on a scale of 10 minutes to hours, includes adjustments of purchases from generating facilities and bringing generators on or off-line to meet demand, and 4) time and inadvertent interchange control, on a scale of hours. Inadvertent interchange control involves frequency changes when financial balances between generating authorities are manipulated. For example, if a regional authority inadvertently provided less energy than it sold, in order to compensate it may run its generators slow while other authorities on the grid run fast. Finally, on a multi-hour scale, the difference between the average frequency and 60 Hz is corrected by purposefully running the grid for a short time at 60.02 or 59.98 Hz. These Time Error Corrections are instigated manually (e.g. by phone to power plants) in the eastern grid and automatically (WATEC) or manually in the western grid. They are initiated when a clock using the 60 Hz line frequency would be inaccurate by +/- 10 s (Eastern Interchange) or  +/- 2 s (Western Interchange). 

NERC proposes to eliminate Time Error Correction, claiming that the only benefit lies in improving the accuracy of 60 Hz clocks, which are becoming outdated by commonly available and more accurate time pieces. In fact, Time Error Correction tends to decrease frequency stability: a study showed that 40% of the excursions below 59.95 Hz (an alarm threshold) in the Eastern Interconnection occurred during periods when the grid was running at 59.98 Hz to correct clocks. 

So, from a LIGO perspective, elimination of Time Error Correction will be beneficial because it will reduce the extent of low frequency excursions that further reduce frequency when the grid is running at 59.98 Hz. During frequency emergencies, e.g. for a large unscheduled shutdown of generation facilities, the frequency is raised by the primary through tertiary controls mentioned above. Time Error Correction plays no role in these emergencies and is only used after the fact to make up for frequency excursions. 

Furthermore, there is a tendency for power authorities to err on the side of over-generation rather than under-generation  because negative consequences begin at smaller deviations on the low frequency side than on the high frequency side. So without time error correction, the average frequency is expected to be slightly higher than 60 Hz. Based on past time error corrections, the new average grid frequency for the Eastern Interconnection (LLO) is expected to be about 60.002 Hz and, for the Western Interconnection (LHO) about 60.0009 Hz. The Crab gravitational wave frequency is expected to be about 59.4 at the beginning of aLIGO, so elimination of Time Error Correction should be neutral or slightly helpful. 

Wall panels were put into place in the acoustic enclosure. Work is coming along pretty smoothly, no big issues so far.

In other news, the laser should be arriving on site early Monday.

- Gerbig working on clean room by H2 PSL
- PSL delivery expected this afternoon is delayed until Monday as freight waits in Seattle
- ICC in BSC8 continues
- Farewell to RobertL, RolfM, RyanL
- Stoneway delivery for EE people
- Much mechanical and electrical activity in H2-ITMY test stand area
- MichaelR transitioning LVEA to laser hazard at 1650

Betsy B., Andres R, Jeff B. 

   We attached the lower wires to the FM in the Staging Building. All masses are now suspended with no apparent gross problems. There is a small amount of pitch in the Top Mass, which we will work on next week.   

The remaining chamber floor and ring sections were brushed although tools remained problematic. As of 2:30 pm, 32 sections had been through first vacuum. The crew will stay in chamber today until the entire upper chamber has been vacuumed so that we can pull the bellows protection and get fibers off the support tubes on Monday.Zack and Randy were able to work on drills a little bit today which is good because the small replacement bearings arrived.

Moved SUS ITMY Sattelite Rack to southwest corner of ISI cleanroom. 

H2 ISI ITMY field cables were re-routed to move excessive wire bundles out of clean room area. 
Swapped out cabling for appropriate lengths (GS-13, T240, L4C, and CPS). 
Ran cabling inside clean room while staying clear of floor.

H2 SUS ITMY sattelite amplifiers were swapped out. Modifications to units are:
1.Replace IC551 & IC552 AD797's with OP27's
2.Replace L101, L201, L301, & L401 with 100 ohm resistors
3.Remove C107, C207, C307, & C407 capacitors.
4.Replace R102, R202, R302, & R402 160k resistors with 121k resistors

R. McCarthy, F. Clara, J. Garner, D. Stone

  We have two failures modes of the interface box,

  the first is on the geophone channels that is characterised by having a DC offset (~4000 counts which i think is ~1volt) on the channel.
The low frequency power spectra is also anomalous. This problem also might be intermittent, as in sometimes goes away when we power cycle.
Richard has this board now and hopefully will come up with a diagnostic.

  the second is in the pressure sensor monitor, some channel slowly (a few seconds) ramp to full voltage, we will pull one of these boxes as soon as the first board gets repaired.

Summary: Magnetic fields from aLIGO electronics were measured in the LVEA test stand racks. Fields from the switching power supplies in the black IO boxes dominated, and were larger than fields from VME crate power supplies, that had produced lines in DARM during iLIGO. I measured magnetic coupling to some of the channels that were working and estimated that the switcher fields were within a factor of 6 of producing features in those particular channels. We may want more of a safety margin than this, so I suggest investigating different power supplies.


I had a look at magnetic fields inside of the aLIGO electronics racks that were set up for the test stands holding the ITMY ISI and SUS, and compared the fields to those in iLIGO racks. I checked several sites but discuss 2 of the most important iLIGO magnetic field coupling sites below. The magnetometer was calibrated in-situ using a coil calibrator. 

Figure 1 compares the magnetic fields in the electronics rack near a test mass controller for aLIGO (ITMY coil driver in test stand racks in the +Y arm of the LVEA) and one for iLIGO (4k ITM controller measured during S6). The 60 Hz fields are significantly greater in the aLIGO setup than in iLIGO. However, this was not due to components in the racks: the fields outside of the test stand racks were about the same as inside and were more than an order of magnitude greater than at typical LVEA sites during science runs because the test stand racks are located near LVEA electrical panels and we are using a lot more current for clean rooms etc. than normal. Otherwise, the spectrum is dominated by broad features from wandering lines produced by the switching power supplies in the black IO expansion chassis discussed below.

Figure 2 shows that the magnetic fields are greater around an aLIGO black box IO expansion chassis than they were near an iLIGO VME crate and are dominated by wandering lines. I opened up h2susb478 and used the magnetometer to search for the source. The fields were much greater right at the Synocean 24V input switching power supply. I confirmed that the fields and source were typical by also checking the TCS black IO box. 

I tried to determine if these fields were deleterious by estimating the coupling of the fields to a few of the channels that were working (OSEM channels). I set up a small coil on the Synocean power supply and generated large 55 Hz and harmonics fields with the coil. I checked that the fields from my dipole coil dropped off with distance about the same way the power supply fields did (ratio of switcher and diagnostic fields varied by less than 3 over a factor of about 5 in distance). 

Figure 3 shows that 2e-6 T/sqrt(Hz) from my injection coil produced about 1 count/sqrt(Hz) on the channels, and so the 2e-8 T/sqrt(Hz) signal from the power supply would be expected to produce 0.01 counts in the channel. While this value is too low to show up on this channel, it is only low by a factor of 6 or so. Considering that 1) other channels may be more susceptible to magnetic coupling, 2) other Synocean power supplies may be noisier, 3) this power supply was not drawing maximum current, and 4) we had line features in iLIGO DARM that were attributed to the quieter switchers in VME crates (here), we may want to consider other power supplies for the IO boxes. 

In a closer look at stationary lines, I saw 10 Hz and 0.5 Hz combs. In the analogous locations during iLIGO, 10 Hz and 1 Hz lines were present, not 0.5 Hz. Possibly something is happening every 2 seconds that happened every second in iLIGO electronics.

Robert S., Richard M.

The split crew continued.

Randy and Zack replaced the conflat on the dome according to M1000355. Kyle, please note for the record that the conflat has been tagged for later inspection. Later in the day, they worked on tooling (aka air drills).

Mark, Chris, Scott H. worked in/around chamber with Carolyn as recorder. Thirty sections were brushed which means that only half of the lower chamber body remains "dirty". Air drills again failed at a rate of ~50%. Bubba and I spent some time looking at a couple drills that had the crew worried because of black residue on the teeth of the chuck (although the metal shroud was clean). Zack disassembled the two drills in question so that we could look at the guts and see what was going on. After examination, Bubba and I agreed that the residue was not likely to have come from the interior of the drill. (The most likely source seems to be particulate that is pulled past the chuck by the vacuum. There is also some possibility that the chuck key could be leaving residue behind when it is used to tighten the brush in place.) I asked Zack to use chuck replacements on any drills that are rebuilt from here on out so that we can keep an eye on the situation.

   Today while paging through some transfer functions we noticed that the low frequency transfer function from stage 1 actuators to Stage 2 CPS does not go to zero, but goes flat below ~100mHz.  This seems like a very strange thing. Last night we took a TF with all of the lockers engaged, 45kgs of extra weight in each corner of the stage 2 keel plate (45*3kgs total extra weight) and with the corner 3 CPS to AA chassis unplugged and shorted.  The attached figure shows the these transfer functions below 0.1Hz and above 0.1Hz transfer functions from the night before when the system was free and fully running

  some things to notice in the plot

  1) locking the system down reduces the low frequency signal (~20X) but there is still alot of signal there (coherence is still 1)
  2) Disconnecting the CPS controllers reduces the signal much more,  reducing the coherence to ~0
  3) We tried disconnecting various CPS sensors in both the locked and unlocked state to see if there was any sensor cross talk, with null results
  
   Very tentative conclusions, either there is pick up from the actuator or actuator drive cable directly to the cps sensor or electronics or the sensors are
actually picking up signal, flexing or motion we aren't sure.

Celine, Hugo and Rich

    Andres R. Jeff K. & Jeff B.

This morning we released H2 SUS ITMY Quad-2 from its stops. All chains and masses are hanging freely and no gross alignment or general problems from the move were observed. The Reaction chain looked good for pitch and roll. The Main chain required 2 turns of fine pitch adjustment to bring it to rough (bubble level and ruler) level.  

I have restored the dust monitor alarm level settings and added directions on how to do so to the h0dustlab, h0dustlvea, h0dustmx, h0dustex, h0dustmy and h0dustey wiki pages. I suspect they were not restored after the power outage.

[Vincent L, Hugo P, Rich M, Celine R]

On Monday, we noticed the following weird reading from the seismometers pods (pressure signals):

• T240 corner 1 & corner 3 read 0
• T240 corner 2 reads ~64 when other pressure reads ~70 (in kPa but we also have figured out that the calibration is wrong)
• L4C corner 3: Pressure reads 0, and diff 18
• GS-13 corner 3: Pressure reads 94 and diff -18

Checking at the pods, we were able to check:

• T240 corner 1 & 3 are actually not returning any voltage => pressure sensors dead
• L4C H3 could not be checked  but the problem moved with the cable => pressure sensor dead
• L4C V3 had a normal readout=> pressure sensor live
• GS-13 V3 had a normal readout => pressure sensor live

At this point, that means that we need 2 T240 and 1 horiz L4C pod replacements.

[Responding to request during morning INS meeting...]
Dust monitor #3 is in the ISI stand cleanroom, #4 is in the SUS stand cleanroom within the LVEA.  The scale of the axes in each plot are different.

Notes:
- At no point did these dust monitors alarm this morning, indicating unacceptable levels.
- Dust peaks in ISI and SUS areas do not seem to be correlated, indicating dirty activity particular to that room, not somewhere else in the LVEA.
- H0:PEM-LVEA_DST_3_5 has been 0 since yesterday afternoon...should probably be inspected.