Detector Watch 2

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6534
Credit: 284710168
RAC: 110462

An entry from the Livingstone

An entry from the Livingstone logs ( Fri Nov 3 18:00:17 2006 UTC )

Quote:

"Fear not the path of truth for the lack of people walking on it."

1415 Locked, seg3550. The microseism is still very high; there are ten-foot waves
in the Gulf, off the Mississippi river delta.

1422 Lost lock, corner station seismic, z-direction, 1-3Hz. This happened
yesterday, too.

1533 Seg3551. It's been windy. We had to wait with the AS trigger engaged for a
truck to remove the scissor lifts from the SEC. He floored it on his way out of
the parking lot; before the paving, we would have lost lock for sure.


Do we all recall the 'thrumming' of the ground when you're a few dunes away from the ocean shore, before you can actually see the breakers of the pounding surf? This is the vibration that is transmitted through the ground which would be part of the microseism as described above. I understand the South-Eastern US is entering a windy season so this will directly affect the interferometer ( IFO ) by the air striking the beam enclosures, but also by driving the waves against the beaches too.

Is there anyone reading this, who lives on the Eastern/Southern US seaboard who can enlighten us to what the general weather expectations are? I saw a mention of an upcoming sports season too in the log a few days ago, what would that be? ( Probably would generate increased road traffic flows..... )

At Hanford a worrying low frequency noise, initially attributed to large fans placed in nearby orchards ( to disperse cold air/frost is my guess ), has been identified:

Quote:
The drop in range between Oct.31 and Nov. 2 was due to seismic noise generated at Energy Northwest. They by-passed the cooling towers on 31 Oct. 2006 at 1740 PST and restored the system on 2 Nov 2006 at 0120 PST. When the opened the bypass valve they could feel the vibration.

There has been some terrific long triple co-incidence periods in the last few days! :-)

Cheers, Mike.

( edit ) Who is 'Energy Northwest'? Sounds power station-ish..

I have made this letter longer than usual because I lack the time to make it shorter ...

... and my other CPU is a Ryzen 5950X :-) Blaise Pascal

Chipper Q
Chipper Q
Joined: 20 Feb 05
Posts: 1540
Credit: 708571
RAC: 0

RE: ( edit ) Who is 'Energy

Quote:
( edit ) Who is 'Energy Northwest'? Sounds power station-ish..


Some history about the Hanford site here. (Wiki page, mentions and links to both LIGO and Energy Northwest, a municipal power company.)

Quote:
Is there anyone reading this, who lives on the Eastern/Southern US seaboard who can enlighten us to what the general weather expectations are? I saw a mention of an upcoming sports season too in the log a few days ago, what would that be? ( Probably would generate increased road traffic flows..... )


Re sports: Not sure how true is was, but I heard that after the World Cup, the US soccer team, before being allowed back into the country, had to prove that they could actually play FOOTBALL :)) Didn't see the log entry, but I'm guessing it was regarding a New Orleans Saints game.
Oops, I checked Livingston's log and found the following entry:

Quote:
You can see that we're in for some stormy weather. Also note that there is a high
degree of correlation between the microseismic noise and the Major League Baseball season.


Probably related to the Superdome, so there'll likely be some noise during the Saint's home games as well.

Re weather: 10-day forecast for Livingston, with current conditions, and links to satellite and radar loops, etc. (scroll down past the ads)

Quote:
There has been some terrific long triple co-incidence periods in the last few days! :-)


S5 Run Statistics 'Daily Locked Statistics' for Hanford on Nov. 2:
H1 Duty cycle 97.1%
H2 Duty cycle 97.6%
Livingston's L1 Science Data Statistics 69.4%

Also found another (Hanford, Nov. 4) entry re: coincident lines:

Quote:
H1,L1 COINCIDENT LINES LIST of Keith Riles; LINE HUNTING
Keith put in a list of ~ 145 lines "coincident" to 10 mHz yesterday
H1 coincident with L1.
We see 6 instances of exact freq match in 95 entrys; diff = 0 mSec.
Is this the right number if random??
The freq resolution for 30 min ffts is ~1/2 mHz so exact nonrandom astrophysics
or devil physics matches SHOULD match at the 1 mHz level.
If random, there are 21 ways to make the table: +-10 and 0 difference;
1 way to be equal: So we expect 95/21 ~ 4.5 exact +- 2+ matches;
6 is thus consistent with this, ie RANDOM lines.
SO, NO obvious devil pathology here.


This sounds very encouraging. What do you think, Mike?
I had to read it several times. It's not encouraging if the coincident lines are expected random happenstance - am I interpreting that correctly?

Chipper Q
Chipper Q
Joined: 20 Feb 05
Posts: 1540
Credit: 708571
RAC: 0

Found a very informative pie

Found a very informative pie chart showing the S5 downtime for H1 since Nov. 14, '05 thru Oct. 26, '06. I was surprised to see that seismic was a cause of downtime only 2.2% of the time, and wind only 1.4%. As you can see, there are about 10 different categories of various causes:
[img]http://groups.msn.com/_Secure/0SQAAACcWcTM11bMGRgOujVbGXBT3LDj6!!CHB8n4y0IAlU073k!WOijsDFeQB7MScPn4griN*wTE8zdgXze2gQsc3uWTi5fe0JxWbvujxk!dpoO9rjHicQ/S5H1DownTime.jpg?dc=4675596471541303107[/img]

The wind at Hanford was picking up, but not too much of a problem for time in Science Mode (H1 stayed locked, but the range dipped slightly):

Quote:
The winds were up sitewide with gusts in the 20-30mphs for a little over an hour this morning. H2 lost lock a few times due to 30+ mph gusts at the corner station...
...
As well, the 4k IR has suffered with large drops due to gusts at the end stations.


S5 Run Statistics 'Daily Locked Statistics' for Hanford on Nov. 3:
H1 Duty cycle 90.7%
H2 Duty cycle 86.4%
Livingston L1 Science Data Statistics 92.4%
There was a mention of increasing microseism at Livingston (elog entry ~8pm local, Nov. 4)...

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6534
Credit: 284710168
RAC: 110462

Here we have an issue worth a

Here we have an issue worth a deeper explanation ( Hanford log Sat Nov 4 01:37:06 2006 UTC ) :

Quote:
H1,L1 COINCIDENT LINES LIST of Keith Riles; LINE HUNTING
Keith put in a list of ~ 145 lines "coincident" to 10 mHz yesterday
H1 coincident with L1.
We see 6 instances of exact freq match in 95 entrys; diff = 0 mSec.
Is this the right number if random??
The freq resolution for 30 min ffts is ~1/2 mHz so exact nonrandom astrophysics
or devil physics matches SHOULD match at the 1 mHz level.
If random, there are 21 ways to make the table: +-10 and 0 difference;
1 way to be equal: So we expect 95/21 ~ 4.5 exact +- 2+ matches;
6 is thus consistent with this, ie RANDOM lines.
SO, NO obvious devil pathology here.


To date I have talked, a bit, about frequency. Now it would be good to discuss how precisely we can measure any given frequency. Let's look at a single cycle of a sine waveform. [ Remember that we can build pretty well any waveform by adding various amounts of various sines together in some recipe. With the resultant curve, we then may rediscover the separate components again by using a Fourier transform. So examining the sine covers alot of ground really. ]

To measure frequency of this we have to do these things:
- find two parts of the wave which correspond ( in the repetitive sense ), and using those.....
- measure the time between those two parts
- take the reciprocal ( if the time interval was measured in seconds ) to get the frequency in Hertz.
Like so:

So what are our limitations here? Taking the reciprocal is the easiest bit. But surely there would be some minimum level of accuracy that we could obtain in identifying either of the two points in their amplitude or their time. The issue becomes the precision with which we could measure the quantity in the vertical axis, whatever that might be, and how finely we could confidently define small time intervals along the horizontal. So now we have:

Those points used to calculate frequency from have now expanded to fuzzy areas of possibilities, the sides of which are generically known as error bars. To improve amplitude resolution we'd build a better detector for that quantity. For the time we'd build a better clock.
Right?

Well, yes we would. However in addition we could cleverly take advantage of the fact that the signal is repetitive ( or assume it to be so ). So for the time axis we would maybe count along as many cycles of the wave that we could, keep only the start time and some end time after some N cycles. While we've gone through many circuits of the wave phase, we still keep the original accuracy of measurement in the time domain. Now the time error per cycle is not delta_t, but delta_t / N .

Quote:
An analogy may help. Down Under we have this terrific annual car race called, generally, Bathurst. As the highlight of the racing year for a certain class of car, everybody wants to win this fast & punishing endurance race. A rev-heads feast in fact ( ie. moi ). Imagine trying to time whether or not a given driver/car combination was performing sufficiently reliably to be trusted to engage in the race, because there are 161 laps over about 6km each over the course of the day - and consistency is the key to success. Well you could time one lap and appraise performance that way - but you could be misled by a crappy driver who fluked a good lap for once in his career or downplay a great driver who was held up by a competitor. Hence it would be better to leave them out for 20 laps, say. You're understanding of their ability to maintain fast laps is improved with those 20 - as you only stopped and started your timer once ( at the start of the 1st lap, and at the end of the 20th ) - as you divide the total time by 20 and hopefully show a better ( average lap ) time than others. The individual laps correspond to signal cycles and we could designate the start/finish line as the recurrent indicator points.

Now suppose I have one signal ( S1 ) whose frequency is truly exactly one ( 1 ) Hertz. That is, every second it runs through one ( 1 ) full cycle, which is the period. Let's pick up another signal ( S2 ) whose true frequency is 1.01 Hertz, thus with a period of 0.99 seconds ( approx ). If I could only measure times no closer than, say, one second apart ( or +/- 0.5 second ), then I couldn't distinguish between S1 and S2 after a single full cycle, in fact it would take 100 cycles. Even more pessimistically suppose my measurement of amplitude was so lousy I could only really tell if the signal was 'up' or 'down' at any given moment, like this:

Now I'm only accurate to within +/- half a cycle ( over the whole sampling time ) because time measurement, and another +/- half a cycle because of amplitude measurement! Not giving up I decide to follow as many cycles of each as I could in order to obtain some distinct/reliable timing difference to distinguish the two ( f = frequency, P = period ) :

f(S1) = 1 Hz

f(S2) = 0.99 Hz

f(S1) - f(S2) = 1.00- 0.99 = 0.01 Hz

and with N = 100 cycles

N * P(S1) = 100 * 1 = 100 ( seconds )

N * P(S2) = 100 * 0.99 = 99 ( seconds )

So now I can say which of the two is which, because the accumulated phase difference between the two has eventually reached my minimum level of discrimination!

Phew! Follow that?

You'll notice a correlation between the desired distinction in frequency ( 0.01 Hz ) and the number of cycles to be sure of achieving that ( 100 ). Yep, they are reciprocals......

Generally you could say that if the desired frequency discrimination is x Hz then then sample time to ensure that is 1/x. The converse is true, if you sample for y seconds then the best you ought to expect for accuracy of frequency is 1/y Hz.

Back to the game:

Quote:
30 min ffts


means Fast Fourier Transforms on 30 minute intervals. In 30 minutes there are 30 * 60 = 1800 seconds. So the best we can do is

1/1800 = 0.0005555 Hz = 0.5555 milliHertz ~ 1/2 mHz

If we have two signals, each of which could be out by that amount, then we'd have to measure them to be at least 2 * 1/2 = 1 mHz apart to be confident they are truly distinct. Now if one looks at signals within a 20 mHz band of some value ( +/- 10 mHz ), then there are 21 distinguishable 'bins' to discuss ( 10 lower than, 10 higher than, and 1 exactly at said value ) like so:

Each of the 95 quoted cases was a pair of frequencies within 10mHz of each other. Put the first of the two, say, at the '0' mark in the bin lineup, then the second is going to lie in any one of them ( and maybe the same one ). Let's be pessimistic and assume there is no relation between the two ( one from Hanford, the other from Livingstone ):

What do we expect from those 95 pairs? Around one in every 21 will lie smack on top of the first ( same bin ), as we randomly sprinkle our 2nd frequency equally throughout said bins within 10mHz of it. With 95 pairs we'd get 95/21 ~ 4.5 of such pairs doing that.

Quote:
Also add in a bit of expected variation in such sprinkling which ultimately goes like the square_root( those_pairs) = square_root( 4.5) ~ 2. It's a long detour to derive this, but if you are curious then investigate the 'random walk' problem - or the 'Gaussian'/'normal' distribution which is the extension of it to large numbers. Basically if N is your expected/probable number then square_root(N) is it's typical spread/variance.

So we get 4.5 +/- 2 ~ 2.5 to 6.5 expected cases for every 95 that you look at. The 6 pairs found were are thus consistent with random occurences. Drats ..... :-(

Cheers, Mike.

( edit ) Based on this type of analysis, we can't say they are definitely not signals of interest ( pulsars ) - simply that the data ( as it is ) does not discriminate between 'actual' pulsar rhythms versus other signals spoofing them ......but heck it's an interesting look at the flavour of the task, and the standard of proof. :-)

( edit ) I hope this discussion emphasises the importance to the project of obtaining good long periods of un-interrupted & co-incident LIGO data. Please don't forget that all our computers at E@H are collectively trawling through the snaps, crackles and pops of the data pipeline looking for the golden needles ..... so keep up the excellent work crunchers! :-)

( edit ) That's an interesting pie there Chipper. The good news is that uptime is at ~75%, and relatively little of the remainder is not understood. It's good to see the design pretty resistant to disturbance from 'external' causes like wind and seismic stuff. Looking good for the future! :-)

I have made this letter longer than usual because I lack the time to make it shorter ...

... and my other CPU is a Ryzen 5950X :-) Blaise Pascal

Chipper Q
Chipper Q
Joined: 20 Feb 05
Posts: 1540
Credit: 708571
RAC: 0

The following Nov. 5 Hanford

The following Nov. 5 Hanford elog entry is fairly typical, except of course for the title:

Quote:
LHO S5 SITUATION: H1 locked ~ 35 Hours; H2 ~ 11. S5 is ONE YEAR OLD.
H1 continues with Segment 2433; Insp Rng 14.6 Mpc;
less flat over last 8 hours
H2 with Segment 1846 Insp Rng 6.8Mpc.
u Seisz ~ .2 u rms; 1-3 Hz EY,EX:.01,.02 u rms.
H1 Cal lines started "drifting" ~ 9 hours ago; =.068 now.
Dave B believes the gamma burst alarm software is non functional:
no entrys for several days to list; I don't remember any lately.
Requires a reload; we opt to do this at a break or tomorrow.
The 35 Hz DARM ERR amp osc is slitely (x 2?) smaller than yesterday;
1 10^-15 m/sqrthz; period ~45 sec.


Congratulations to the LSC and LIGO scientists on reaching this S5 milestone!!
S5 Run Statistics 'Daily Locked Statistics' for Hanford on Nov. 4:
H1 Duty cycle 99.3% [Wow!!]
H2 Duty cycle 87.1%
Livingston L1 Science Data Statistics 93.5%
Interesting to note the there's about a 3 Mpc difference in the inspiral range between L1 and H1...

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6534
Credit: 284710168
RAC: 110462

I followed up the Energy

I followed up the Energy Northwest reference, thank you Chipper! It's the Columbia Power Station:

[ The Fast Flux Test Facility I gather is an all but defunct experimental nuclear reactor... ]

and while we're Google Earthing, here's that train line that rattles Livingstone:

At Hanford there have been:

- an anniversary, the S5 start!
- unexpected injections.
- an earthquake, 2.6 magnitude @ 15.7km depth under Portland Oregon 05:34:36 UTC 06/11/2006.

At Livingstone there have been:

- an anniversary, the S5 start!
- trains ... :-)

Now this is what I call a State Vector plot:

Clearly November has been a kinder month than the last. Hoping there be many more like it! :-)

Cheers, Mike.

( edit )

Quote:
Interesting to note the there's about a 3 Mpc difference in the inspiral range between L1 and H1...


I missed that Chipper. Well spotted. I'll have a dig around, maybe see why...

( edit ) also 35 hours of lock gives 35 * 60 * 60 = 126,000 seconds. Best possible frequency discrimination is thus ~ 10 uHz ( u = micro = 10^(-6) = 1 millionth ). That is you could tell the difference between about 1.00000 Hz and 1.00001 Hz, as after ~ 100,000 seconds the first signal has one less full cycle/phase accumulation than the second. This assumes one can be sure of counting/discriminating to within a full cycle over that time segment.

I have made this letter longer than usual because I lack the time to make it shorter ...

... and my other CPU is a Ryzen 5950X :-) Blaise Pascal

Chipper Q
Chipper Q
Joined: 20 Feb 05
Posts: 1540
Credit: 708571
RAC: 0

30, 40, and eventually 50

30, 40, and eventually 50 mile per hour winds at Hanford. Here's a bird's-eye-view of conditions at the time of this post:

Some trouble with ETMy again, good recap in the Owl Shift Summary (Nov. 6, 8:30am local) entry:

Quote:

It was windy throughout the night.
Around half past 7:00 (UTC) the wind speed meter at Mx, Ex started showing figures that exceeds 30MpH quite often, and a few hours later some started showing more than 40MpH. The two IFOs' lock has been unstable most of the time.

When the H2 dropped out of lock the last time around 13:30, we saw a bump on the H2:LSC-AS_Q spctrum,from around 10Hz to 400Hz, and this problem was seen a few weeks ago, after the big earthquake around Mt.Rainear. They found some problems in Etmy, so in order to investigate this, first ETMy of H2 is undamped, all pitch, yaw, position modes were observed, but they didn't seem as bad as they were when they had the same problem before, around 24th of October. So, later every main optic of the H2 was undamped. The investigation is still going on.


And here's a treat from Livingston – 1 year's range, laser power, and seismic (a couple ranges) at a glance [elog entry with description follows plots]:
[img]http://groups.msn.com/_Secure/0UQDdAo4ZPTSCbTA1LIfrmr*tyVnTv*Cs0LScwWhT4bIFFqSn6vvSFZ7TsMspDWmF6JIY9pysZ*Z105SpwKejiegn87o1ynCX7u8xFLW80f1qcdpW2mrWOvaHaxU7op6V/1YearRangeAndSeismic.jpg?dc=4675596757585929361[/img]

Quote:
One year trend of BNS Range, Laser Power, uSeism
The attached plot shows one year worth of data. To select the night time data I imposed
a cut for LVEA-X rms velocity in 1-3 Hz seismic band <0.09 um/s. The BNS range is plotted for various power levels using different colors (the colors in the legend did not come out so I added letters for colors to the legend: g-green, r-red, c-cayn, b-blue). I used the MC transmitted QPD sum as measure of power into interferometer. I believe the calibration of this PD has not changed in last year. It is not clear if we have more upconversions now than we had in the past. The uSeism is still on the rise and, as several people mentioned before, it will peak around January.
Chipper Q
Chipper Q
Joined: 20 Feb 05
Posts: 1540
Credit: 708571
RAC: 0

Not too much time in Science

Not too much time in Science Mode lately. With Hanford still having weather related difficulties, at Livingston they used the opportunity (or technically lack thereof, with regard to collecting coincident data) to perform various alignment checks and troubleshooting. In both elogs, there were many charts and plots comparing different noise factors (e.g., wind or microseism) to trends in the inspiral range, and also comparing previous conditions of alignment control data (before events like quakes or burps) with current conditions.

A couple interesting issues remain; one is the inspiral range for L1 (needs be a couple Mpc greater compared to H1, as they're both 4k IFOs), although different levels of microseism (or differences in other environmental noise) between the two observatories would make that an unfair comparison, and the other issue is the suspension for the optics, and how best to fine tune the alignments for the greatest detection range under the varying conditions.

Haven't had the chance to check on the details, but I think there are three stages of commissioning for the suspensions, known as 'initial', 'enhanced', and 'advanced' LIGO...

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6534
Credit: 284710168
RAC: 110462

OK, gravity detectives....

OK, gravity detectives.... here's a real treat! :-)

Mike Landry kindly made himself available to answer some questions from myself & Chipper as regards the operational side of LIGO. Mike is co-Chair of the LIGO Scientific Collaboration's Continuous Wave Search Group, one of the experimental experts who works at LHO and he spends a lot of time in the Hanford control room.

Q - What is the average number of circuits of a photon in the Fabry Perot resonance part of the interferometer ?

A - It's a continuous process, but on average a photon bounces about 200 times in the arms, making it effectively an 800km long arm, instead of a 4km one. This amplifies the phase delay of one arm relative to the other, so you're sensitive to a smaller gravitational wave (h = deltaL /L, with L now effectively larger).

Q - Do the mirrors 'roll' ie. rotate around an axis through the centre of, but normal to, the face ?

A - Yep. The large optics have a roll mode at about 18Hz, (root 2 * their vertical bounce mode). You can see this mode as a peak in the AS_Q and DARM_ERR spectra in the elog. If you had perfect aligment, the length degree-of-freedom would be insensitive to this mode. However, nothing is perfect, and the roll mode makes length noise, and that gets picked up in the error signal of the interferometer.

Q - Does FB as in 'Thus far in S5, we've had a couple of periods where we (1)had ALL optics undamped for more than 45min, and (2)had no FB issues (so, power outtages, don't count!)', mean 'forward bias' ?

A - FB here means "FrameBuilder". This is the CPU responsible for assembling all relevant channels (both interferometer ones, and environmental/facility ones, about 10,000 channels per interferometer, some fast, some slow, ~10Mb/s per interferometer) into a file structure called "frames", and writing them to disk. This is the base unit for gravitational wave analyses; raw frames. From there we make other data products, to simplify analyses, but those are made offline.

Q - Why are some plots ( spectral density, the vertical axis ) in 'per _square root_ Hertz' ? Something to do with sample length ? I think I understand per Hertz!

A - This comes out of the history of the hardware of frequency analysis: in a lab, you'll tend to use a spectrum analyzer, and plug signals into it, and determine how much power is in a given frequency band. The units are then Power/Hz, or if you have calibrated your channel, Watts/Hz. We tend to work with amplitude instead of power, and as power goes as the square of amplitude, our frequency spectra are then in units of Amplitude/squareroot(Hz), or if they are calibrated, Metres/squareroot(Hz), or perhaps Strain/squareroot(Hz).

Q - How many FOMs are there that aren't automatically e-logged ?

A - While there are many other offline analyses that are used to characterize how the machines are performing, making use of many of the 10,000 channels/interferometer noted above, all Figures of Merit (FOM plots) are by definition all logged automatically to the elog: FOM1-3, DARM_ERR and AS_Q, and the site overviews.

Q - Which FOM, or combination of FOMs, do you refer to most often, in an average day at the observatory ?

A - It may depend on how the machine is functioning. FOM1 is a very good way to assess how the machine is working. The binary inspiral range (i.e. how far our could a given interferometer detect a 1.4/1.4 Msun binary neutron star system, with a signal-to-noise ratio of 8, angle averaged over all sky positions and all binary orientations?) is a key metric of performance because it is sensitive to a large range of frequencies, right around the sweet spot of the interferometers (~80-200Hz). The number, say, "15Mpc", gives you a lot of feel for how well the machine is, and the hash on the plot tell you if it is glitching. On windy day, you might end up looking at glitch monitors (FOM2, FOM3) more in order to see trends in burst behaviour on interferometers. All FOMs are scrutinized each day however. The noise analyses (AS_Q and DARM_ERR fourier series) are immediate feedback on the frequency components existing in a machine, very key information. Is a test mass mode ringing up? e.g. do I see an 18Hz peak popping up in the last few minutes (evidence of a roll mode, noted above - an optic has begun to roll).

Q - What's the total number of sensors, in the entire facility, for which you can generate specialized FOMs ? (ballpark, if it's 100s or 1000s)

A - You can make time trends or fourier analyses of virtually any of the ~10,000 channels per interferometer; we tend to refer to FOMs as those special channels that tell us estimates of how the machine is function *with respect to astrophysical estimates*, e.g. FOM2, top RHS: "how long would it take us, with the current noise in the machine, to beat the Crab spindown limit?" You cast the noise in terms of some astrophysical goal or metric. But otherwise, we tend to look at many other channels either in the time-domain, or the frequency domain, using software tools, and not refer to them as FOMs.

Q - What do you refer to most often when looking at the 'Site Overview' (during an average day) ?

A - To zeroeth order, you look for red: you're looking for deviations from good behaviour, which gets flagged: green for good, red for bad or changed. You want to ensure things are functioning well. When you're on shift, you have a certain mentality of "not on my watch". You don't want anything to go wrong and you want the machines to function as well as possible. The site overview has lots of info, some of it for specialists, e.g. vacuum specialists may monitor specific measurement points to see trends in pressure, if they suspect a meter or a seal problem. Another important site overview element is the "Stat" region - have any running parameters changed recently? They will then be flagged in red.

Q - How many people does it take to keep the observatory up and running, usually, during any shift ?

A - A minimum complement is two: one operator to run both machines, one scimon to ensure the data be written and the astrophysical outputs of the machine are of sufficiently good quality. We run graveyard ("owl"!) shifts this way. You don't want to run too long this way though - there is more intervention generally required than two people can do... you could only last a few contiguous shifts this way.

Q - Recently there was a notification from re: RXTE crew and window of opportunity for a concerted effort to observe Sco X-1. Are there notifications like this for other satellite observatories ? How many, which ones ? About how often ?

A - RXTE is the only electromagnetic observatory with which we've coordinated specific observing time. We hope to make several such coincident runs with them. However, you can see at the bottom right of the site overview a little "GammaRayBurst" icon. This flashes red when we get an email from the Swift of HETE space-born gamma ray telescopes that a GRB has been detected in the last few seconds or minutes. If our machines are operating at the time of the notification, and we were contemplating mucking with the machine, we will hold off. For example, I will not take the interferometers out to calibrate them, if a GRB has been detected within the last hour.

Of course, we coordinate our runs all the time with other gravitational wave detectors, like the GEO and Tama observatories, and in the future, Virgo. We also coordinate with Allegro and the other IGEC-2 bar detectors.

Q - I noticed one day that microseism activity made keeping a lock impossible all day long, and I thought it was particularly interesting that the activity appeared to be increasing (~10am local, so after morning rush hour). Are there any geologists taking advantage of all the seismometer data ? Have you noticed many lock losses that preceded (by minutes or seconds) events qualifying as quakes elsewhere in the world ?

A - The microseism fundamental mode is at a period of about 7s, or 0.15Hz, so it is distinct from cultural noise like trucks and commuting. The microseism is due to ocean waves beating against the continental shelf, and it rises in the winter owing to wave activity.... you'll definitely see this hampering us more in Dec/Jan. Quakes, typically at higher frequencies like 3-10Hz, hit all the time and are a distinct problem in maintining lock over multi-day timescales. Close quakes can be trouble, like the extremely shallow and close Mt Rainier quake of October:

Geologists don't use our data now, although they can, and they may in the future. I think they just have better (seismic) instruments than we do, and probably do a better job arranging them for their purposes. Mostly we want to *decouple* our interferometers from the earth; we want to simulate free fall in space, force-free test masses unadultered by acoustic or seismic coupling, so the detectors should be insensitive to these effects. The seismometers we use are good ones, but I think the geologists have better, or make better arrangements, given their interests.

Q - LIGO seem to be sensitive to noise from a plane's propeller, but what about noise from sonic booms? Would jet aircraft cause such issues with regard to their powerplants?

A - Loud noises are bad; prop planes, jet planes, sonic booms. They shake optics external to the interferometer, are picked up on the light as Doppler shifts, sent back into the interferometer and amplified in the arms, and then sent out the gravity wave port to be registered by our computers. However, we have done much to mitigate acoustic coupling in our machines. Gravity wave photodiodes now live under large acoustic baffles, like igloos, to reduce coupling. We think we're in pretty good shape for acoustic coupling now.

Q - I ( Mike ) have purchased Peter Saulson's "Fundamentals of Interferometric Gravitational Wave Detectors", and M Vaughan's "The Fabry-Perot Interferometer (Series on Optics and Optoelectronics)". Could you recommend any others ? I love a good textbook - certainly happy to spend on good ones - and I don't fear depth, though I probably should.... :-)

A - Well you have some good material on interferometers and optics there, perhaps you want to complement them with a good book on relativity?

Bernard Schutz, "A first course in relativity" is considered an excellent starting place.

A book I'd recommend because of its breadth (not because of its direct applicability to LIGO) but may be hard to find (I don't know), is, "300 Years of Gravitation", edited by Hawking and Israel. I'd known about this for a long time but only started reading it on recent travel. It is part historical retrospective (giants of field reflect on Principia anniversary), which makes for fascinating reading.

So there you have it! A terrific insider's look at LIGO operations. Many thanks, again, to Mike Landry for making himself available. :-)

Cheers, Mike.

I have made this letter longer than usual because I lack the time to make it shorter ...

... and my other CPU is a Ryzen 5950X :-) Blaise Pascal

Chipper Q
Chipper Q
Joined: 20 Feb 05
Posts: 1540
Credit: 708571
RAC: 0

Well, there's been some

Well, there's been some ongoing commissioning work with H1, but it will hopefully be back in Science Mode 'tonight', which should be around the time of this post (+- a few hours), have to wait for the next elog entries. Fortunately the wind's no longer a problem. The work is on the ISCT4 float system. In the process of trying to figure out what ISCT4 is, I found this LLO 4k Optical Layout (Note ISCT4 in the lower right-hand corner):
[img]http://groups.msn.com/_Secure/0TwAAAM8Y8PZsXleLMb5sL5*2sKwIOlONEZLTZh73TSRPZrZ8vSjk3rZ!Z1DjAJwgsKrLWh0HoGak8x3Ph4*jjGq4KwTLq5Su4vUSKWEQTk1qEsCOIb8T3g/LLO4kOpticalLayout.jpg?dc=4675597045584515991[/img]
Lots of work being done at Livingston checking on microseism upconversion, among other things, and they're currently in Science Mode with an inspiral range of ~12.5 Mpc.
Wow! Awesome Q&A!! Thanks Mike Hewson, for putting that together, and special thanks (again) to Mike Landry!!

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.