Detector Watch S6 V2

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6,580
Credit: 307,755,034
RAC: 187,595
Topic 194579

This thread continues from this one. I'll repeat my disclaimer so that no one mistakes me for a real physicist! :-)

Quote:
This is unofficial and represents my views and understanding alone, but please feel free to add to the thread anything of interest. I read the Hanford and Livingston online logs - which you can too if you like - as a guest. Just use the username 'reader' with password 'readonly' when selecting the detector log link.


Now .... when Hanford is not being jiggled by rock hauling trucks, school tour groups, renovators, earthquakes, stray thuds, winds and maintenance activities - it's been collecting science data! :-)

The 60Hz noise with harmonics at 120 Hz and 180 Hz are also getting a mention. Along the lines of the range drops with temperature, there seems to be a pretty good time correlation between the temperature in the area where the Output Mode Cleaner is and the control of it's suspension.

Now remember "the table on which the optics sit ( LVEA - ISCT4 ). It sits on a cushion of air, so to speak, in order to reduce vibration" was problematic ? It has been inspected :

Quote:
We then went out to look at ISCT4. Nothing looked obviously wrong, except that a spider had crawled up onto the table, spread a web over an unused optic near ASPD3, and died. We removed said dead spider, looked for clipping or scattering, but did not find anything.


I wonder what the force magnitude of the average footfall of a spider is? This sort of thing wouldn't happen if in vacuum .... except if a lady called Ripley starts doing LIGO shifts. :-)

Livingston has worse, Atlantic storms pounding waves onto the shore I think. Little joy for science .... few segments collected. I've looked at a paper that discusses changes to the OMC geometry to improve it's performance. The following is something similiar to the present arrangement, and receives the light from the dark port of the interferometer, but after the pickoff's for some of the sidebands that control the IFO.

This is the "cat's cradle" that I've referred to earlier, but is actually called a 'bowtie' arrangement and has 4 mirrors for the resonant cavity compared to 3 for the Input Mode Cleaner. The idea is to remove all other frequencies besides the base/carrier frequency - cut out those sidebands mentioned earlier that are used to control the IFO, plus any 'junk' light hanging around. This is to get the photodiode output to represent, as much as possible, the phase changes due to differential arm length altering. We are/will analyse the signals from those labelled DCPD1 & 2.

The light comes in from the bottom, hits the steering mirror, goes though the back of mirror 1, to the face of mirror 2, to mirror 3, then mirror 4 and back to mirror 1. This is the primary cavity. Same racetrack idea as the input mode cleaner. Recall that mirrors can be designed with specific desired amounts of reflection and transmission. Also light within beams that cross don't affect each other. The science readout is derived from the beam that goes through mirror 3. The near field and far field photodetectors are used for coarse and fine alignment respectively of the steering mirror position I think.

I tried to follow the discussion of why 4 versus 3 mirrors in the cavity. As far as I can tell the basic issue is that each mirror changes the light phase by 180 degrees on reflection. So 3 mirrors gives 3 x 180 = 540 ie. equivalent to 180. Whereas 4 mirrors gives 4 x 180 = 720 ie. equivalent to zero. This changes the properties of resonance in ways I couldn't follow ... making a 4 mirror cavity better in certain respects than a 3 mirror cavity. Perhaps another day I might be able to make more lucid comments on this, but don't hold your breath in the meantime. :-)

Cheers, Mike.

( edit ) Oh dear, the rabbits do have a problem.

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

ML1
ML1
Joined: 20 Feb 05
Posts: 347
Credit: 86,484,412
RAC: 2,763

Detector Watch S6 V2

You've mentioned wavefronts on various occasions...

Your diagram there suggests that the photodetectors are at a small angle so as to deflect any reflections from them away from the incoming beam.

Doesn't that 'blur' the interferometry wavefront they are supposed to be detecting?

How can you be sure that the surface and active volume of the photodetector material is 'flat' enough to precisely detect any phase shifts?

Thanks for the good commentry as ever! So they have rabbits that leave trails that glow in the dark!!!

Cheers,
Martin

See new freedom: Mageia Linux
Take a look for yourself: Linux Format
The Future is what We all make IT (GPLv3)

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6,580
Credit: 307,755,034
RAC: 187,595

RE: You've mentioned

Message 95176 in response to message 95175

Quote:

You've mentioned wavefronts on various occasions...

Your diagram there suggests that the photodetectors are at a small angle so as to deflect any reflections from them away from the incoming beam.

Doesn't that 'blur' the interferometry wavefront they are supposed to be detecting?


No. One needs to distinguish photons that are used for various control purposes from those that give us the gravitational wave sensing.

We are comparing those two paths in the arms for interference. We try to ensure light is in the same phase before the splitter that sends them to one or other arm. For those that return to exit via the anti-symmetric port, from there on they are all on the same path. Regardless of what interference has or hasn't occurred between the arms the photons then suffer the same phase changes onwards together after leaving the dark port. So the phase difference ( for a given frequency ) is fixed beyond the recombination of light from the arms. Moving/realigning the post-OMC PD's positions won't change any relative phase shift that has already occurred.

So Barry The Photon might have been born, laser synchronised in phase, together with Charles The Light Particle : but their experiences may only differ if Barry took the X arm while Charles took the Y. Or vice-versa. What particularly matters for our GW signal photodetectors is that both Charles' and Barry's relative phases are compared. There's quantum mechanical probability amplitudes and whatnot here determining whether an electron get's shoved to move along in our subsequent circuits, so we want Charles and Barry to combine votes on that. Plus it's good to have PD's that like to soak up as many photons of interest as possible, hence we get as many Barry/Charles pairs as we can.

A 'wavefront' answers the question : if I start at some given place, what possible future positions ( some time, some place ) exists for me at some particular chosen phase value? It will be some 2D surface(s). If I chose a different phase value I'll get in general a different surface. So I can speak of the '90 degree' wavefront, or the '2.5578 radians' wavefront. Admittedly phase is often quoted in 'angle' language, which can be confusing as polarisation ( another topic entirely ) is quoted as angle too. Best to think of phase as a pure number that ranges from zero to some maximum value which you can't exceed - you start back at zero instead! You can't wind around under zero either, as you then get put up to that maximum. This is termed modulo arithmetic. Note that although I can subtract/add a full cycle of phase and get the 'same' phase, for some specific photon, that then means some earlier/later time and closer/further distance from source. So wavefront language is a way of discussing the propagation in time and space of particles, like photons, that have underlying cyclic phase character.

So if I am Edward the Electron inside the PD crystal ( .... minding my own business .... circulating around The Nucleus .... shooting the breeze and downing wine spritzers with the other electrons in the 'hood .... ) and then Barry and Charles arrive in possibly different phases at my position at a particular time. I'll compare them and decide accordingly to absorb energy and hiss off. Or not and continue to hang around. Note that Barry's wavefronts will in general be different from Charles', but it is possible that they may meet Edward in phase ie. their wavefronts of that phase happen to co-incide at Edward's place. Another way to express matters is that generally when they arrive the photons will be on wavefronts of different phase. If Eddy was asleep when they turned up, and they were compared elsewhere deeper in the crystal that's OK. Or if someone else got to them before Ed too. Or if Barry and Charles entered laterally displaced from Edward .... etc. What matters is that some electron in the active volume does the job.

Quote:
How can you be sure that the surface and active volume of the photodetector material is 'flat' enough to precisely detect any phase shifts?


Well the situation is different for those sideband photons we use to control the interferometer geometry in realtime, than it is for the likes of Bazza and Chuck who are 'science' photons. I think I may have been a bit misleading in an earlier post. In any case the wave front sensors used for IFO length sensing do have to be concerned with whether or not a wavefront ( read 'surface of constant phase' ) hits square on. That is so that one can have a single position/meaning for the 'end' of an arm or cavity. That's why there are many pickoff points around the interferometer : take a sample of the traffic, select photons of interest ( frequency, phase, polarity ... ) from that, count how many you have, repeat continuously.

It is a matter of ( considerable!!! ) convenience that we use the same laser, vacuum space, optics etc to host/house the control elements. Michelson & Morley did it all by hand.

Quote:
Thanks for the good commentry as ever! So they have rabbits that leave trails that glow in the dark!!!


Thank you! If the rabbit and/or the poo glows -> run away!

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

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6,580
Credit: 307,755,034
RAC: 187,595

Hanford has been having a lot

Hanford has been having a lot of trouble locking. There's been renovations continuing which is either preventing or upsetting stability. A bake oven is being installed - this will be used in onsite component manufacture, not pastries. :-)

Livingston is doing better, but the Atlantic Ocean rumbling and a variety of earthquakes have conspired.

Now both sites have been mentioning, here and there, issues with OMC ( output mode cleaner ) and WFS ( wave front sensors ). It seems the issue is one of control. To that topic I've done a spot of reading, found this very useful article, and will share some interesting points gleaned therein .....

This is a good diagram of the Advanced Ligo interferometer layout :

where

ETM = End Test Mass
ITM = Input Test Mass
PRM = Power Recycling Mirror
SRM = Signal Recycling Mirror
BS = Beam Splitter

So the laser light comes from the left side, passes through the back of the PRM, hits and thus divides at the BS, passes to one or other of the ITM's and onwards to the ETM's 4km away. Returning light may bounce off an ITM and thus return back up 4km .... etc for many laps. Some will return through the ITM to the BS and travel to the PRM or the SRM. Importantly while LA and LB are 4km, the 100 or so laps within the ITM - ETM cavities ( Fabry-Perot or FP ) effectively 'folds' a much longer path ( 100 * 2 * 4km = 800km ) into the IFO facility. So whatever phase change a photon undergoes with one lap ( there and back ITM ETM ) of the FP cavity is multiplied by the number of laps. This amplifies the differences of phase between the two arms when a gravitational wave passes.

In the normal or expected mode of operating we want little light coming through to the SRM ( ie. dark port or science signal ), thus the bulk of the light will go back to the PRM. The PRM doesn't do any length sensing or GW detecting, but is there to keep those photons in the deal. Shame to waste them. For AdLigo the aim is to achieve about 1MW in the arms. Yes, that's one megaWatt of photon power circulating in the FP cavities! That's enough for about a thousand domestic vacuum cleaners. I don't know about you, but that impresses me! :-)

In principle it would be 'possible' ( not ruled out ) to do GW detection with a pure Michelson arrangement. Just remove the PRM, SRM and the ITM's. In theory you could use the dark port reading, but we'll do oh so much better with these design enhancements. So what's the SRM for then? It can't be for power per se as it's at the dark port. It will be for 'tuning' the IFO. Enhanced LIGO ( ie. currently ) doesn't have this, so we'll leave that aside. However the OMC presently is in place b/w the BS and the SRM position in the above diagram.

It's important to note that what we deem as the configuration to make the dark port 'dark' only refers to the base/carrier frequency produced by the laser ( 1064 nm ). The sideband photons mentioned earlier are at a different frequency and they will generally come out the dark port easily. As we want them to, because we use them to do adjustments. What is it we are adjusting? Basically mirror positions and angles, plus the laser frequency too.

Generally what is used to adjust the optical components to a fine degree is the following. Get a small magnet. Glue it to some aspect of the optic of interest in a position along the line that you want move the optic. Nearby, but not touching, you place a small conducting coil, suitably oriented and fixed to some adjacent mount. So by passing a small current through that coil, a tiny magnetic field will be generated that in turn pulls on the magnet and hence the optic. Plus if you reverse the current direction the field sense is flipped and a push becomes a pull ( and vice-versa ). In addition small lasers are shot at the magnets in a way that the magnet will block more or less light depending on where the magnet is currently sitting. Put a photodiode on the other side to catch the light and hence you have a way of seeing by how much the beam is being occluded/eclipsed by it. A position sensor in fact.

So you can see that one could set a certain coil current to place the mirror, and receive a certain PD current to gauge the effect of that. That of itself could well be sufficient to control a small IFO. But at 4km away one mirror looks awfully small! So the tiniest of misalignments in the direction that a mirror is pointing is disastrous. More than about 10^(-8) radians in fact, which is fractions of a millionth of a degree of arc. And not just for a photon to make one FP lap, but to keep returning ~ 100 times over. So angular control is critical, and mostly so for the arms.

Now to the Wave Front Sensors. There are several of these at certain points which operate to sample different photon frequencies. It's a bit like a remote control toy aeroplane, which has a number of distinct radio channels, each for a certain type of control surface movement ( ailerons, rudder, throttle ... ). A technique loosely called quadrature is used. Imagine we're looking at a wall, with a light shining on it from far away. We have some central target area that we want the light to hit as accurately as possible. How would you automate this? Try drawing a big crosshair on the wall, that is : perpendicular intersecting lines, centered on the spot of best alignment. Now get four people, each of which look only at one quadrant. So Mike looks at the upper/right one, Martin the upper/left, Magic the lower/right and Tullio the lower/left.

We each regularly report the brightness that we see only on the part of the wall assigned - say we score it from 0 to 10. Each of those scores is simultaneously read and transmitted to whom-so-ever is controlling the light pointing. Could you deduce the condition that decides if the beam is hitting center? Certainly! When each watcher reports the same brightness score ( or close enough to some tolerance ). Could we construct a method to 'steer' the beam toward that ideal state? Yes again! The general idea is 'steer away from the quadrant(s) that report the brighter numbers and toward the quadrant(s) that report the dimmer numbers'. It is somewhat rather more involved than that, but you get the idea.

I'm not sure how this quadrature is currently physically achieved. It can be achieved by having a prism with an edge that intersects the beam line, the faces either side of that edge being reflective. This acts like a knife to split the beam to, say, left and right sides. Each of those are then subject to another similiar prism, slicing them to up and down. What emerges is four beams each with an amount of light corresponding to how well the beams hit the prisms along their edges. If the center of a beam is to one side of an edge more light will reflect to that side and less to the other ( the beam has width ). Plonk a photodiode at each of the four emerging beams and you have Mike/Martin/Magic/Tullio. It could be like the Gravity Probe B guide star alignment system :

or even as a single photodiode :

Feed the PD currents ( after analog to digital conversion ) as inputs to an algorithm ( matrices and stuff ... ) to then yield numbers which then go back ( via digital to analog conversion ) as the coil current that adjusts an optic. You now have a control loop to auto-steer the mirrors. You can't rely on the occulting magnet method to determine the best angle to steer a mirror. You use the occulting to know what it's position/angle is, but you need information fed back from 4 km away to know if it's a good one.

There's plenty more to come from that paper. Stay tuned. But 'tis enough for today. :-)

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

ML1
ML1
Joined: 20 Feb 05
Posts: 347
Credit: 86,484,412
RAC: 2,763

RE: ... while LA and LB are

Message 95178 in response to message 95177

Quote:
... while LA and LB are 4km, the 100 or so laps within the ITM - ETM cavities ( Fabry-Perot or FP ) effectively 'folds' a much longer path ( 100 * 2 * 4km = 800km ) into the IFO facility. So whatever phase change a photon undergoes with one lap ( there and back ITM ETM ) of the FP cavity is multiplied by the number of laps. This amplifies the differences of phase between the two arms when a gravitational wave passes. ...


The reflectivity and transmittivities make for interesting numbers:

Transmission and reflection coefficients (40m prototype, Advanced LIGO):

t1^2 = t3^2: 0.0050000, 0.0050000
r1^2 = r3^2: 0.9949625, 0.9949625
t2^2 = t4^2: 0.0000150, 0.0000150
r2^2 = r4^2: 0.9999625, 0.9999475

One thought is that the light from the photons from just one or a few laps should dominate (swamp) the signal and the that 100-laps photons would be so few as to make for only a very weak signal. However...

With those 99.5% reflection / 0.5% transmission, the one-lap light only makes up at most 0.5% of the signal, but you still get a steady summation of the multiple laps until at 100 laps the light comes back from a level of 60.3% to give 0.3% light return. Hence, is not the interferometer 'desensitised' by all the low-numbers-of-laps returned light?

Or is there something more clever going on with the arms acting as a Fabry-Perot cavity?... But is that effect not back at the beam splitter?

Regards,
Martin

See new freedom: Mageia Linux
Take a look for yourself: Linux Format
The Future is what We all make IT (GPLv3)

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6,580
Credit: 307,755,034
RAC: 187,595

RE: The reflectivity and

Message 95179 in response to message 95178

Quote:

The reflectivity and transmittivities make for interesting numbers:

Transmission and reflection coefficients (40m prototype, Advanced LIGO):

t1^2 = t3^2: 0.0050000, 0.0050000
r1^2 = r3^2: 0.9949625, 0.9949625
t2^2 = t4^2: 0.0000150, 0.0000150
r2^2 = r4^2: 0.9999625, 0.9999475


Yup. If you take, say r1 and t1 together then there is precious little spare change from unity/one. That's how little absorption there is per bounce. I reckon those mirrors must have cost a fortune, plus much brow sweat, to produce. I wouldn't want to be the one that trips and falls headfirst into it ... :-)

Quote:

One thought is that the light from the photons from just one or a few laps should dominate (swamp) the signal and the that 100-laps photons would be so few as to make for only a very weak signal. However...

With those 99.5% reflection / 0.5% transmission, the one-lap light only makes up at most 0.5% of the signal, but you still get a steady summation of the multiple laps until at 100 laps the light comes back from a level of 60.3% to give 0.3% light return. Hence, is not the interferometer 'desensitised' by all the low-numbers-of-laps returned light?


That's one reason why the IFO's sweet spot is down around 100 - 200Hz. What photons that are held for longer in the FP's are probing the higher frequency end of the GW band. So they take more risk of absorptions, deflection from stray molecules in the vacuum, change of GW profile during transit, tax audits, Fridays .... :-)

If AdLigo reaches 1MW in the arms it would have done so with a 250W laser. And everytime you lose lock much falls out - the dark port isn't dark - so you have to pump back up. The more multi-lap a photon is the more power per photon it represents, by accounting for the fraction that didn't make it ( there's a gruesome analogy with WWII flight crews here ). You can't get around it, as by definition all 50 lap photons were 5 lap photons at one time.

Quote:
Or is there something more clever going on with the arms acting as a Fabry-Perot cavity?


Yes, that being the delicate dynamic control of the cavity to acquire and then stay on resonance.

Quote:
But is that effect not back at the beam splitter?


I've left the BS story for another day. But if it's at the right position and angle it should be left alone thereafter, not partaking much in on-the-fly adjustments.

I guess another way of looking at this is the FP cavity's finesse. A graph of the light intensity vs frequency for the FP ought have a sharp peak. Finesse - or 'Q factor' - is a measure of the height over the width of that peak. But that also means finesse represents the stored power divided by the rate of it's loss out of the cavity. Any photon not on the peak frequency ( thus best wavelength for the current cavity dimensions, so it's phase upon return from a lap matches other outgoing photons ) is going to successively risk missing the cut. By definition the cavity is not quite resonant for those guys, but it will still take some laps before their personal phase shift drifts too far. The further from the peak frequency a photon is, the fewer laps that will take, and the sooner that will happen.

I can see that you've picked up on how fascinating are the nuances of this LIGO gadget!! :-)

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

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6,580
Credit: 307,755,034
RAC: 187,595

Both Hanford and Livingston

Both Hanford and Livingston have had their share of woes this last week or so. I am reminded of the old rhyme about 'for want of a nail a horseshoe was lost, for want of a horseshoe a horse was lost, ...'.

Livingston can lock, but at around 8W only, getting bumped out with seismic noise - generally from earthquakes around the world and the pounding of the Atlantic waves against the continent. There is/was a tricky problem of timing between circuits reading out the data and handing them on to buffers. 'wait' cycles were being inserted sometimes, thus introducing variable delays and now it appears one circuit is giving a rising edge when another is expecting a falling edge from it. So for the moment :

Quote:

An inverted (or not inverted when it is expected to be) 65kHz clock would shift the OMC timing by about 8us, which might be enough to explain our timing problems. We will need a bunch of non-inverting 2-pin lemo cables to test this... Radar?

We left the timing as yesterday, but changed the OMC code to use the double-buffer and to write only on even cycles. It should write on odd cycles, but since the delay is 2 cycles rather than 1, samples must be written on even cycles to arrive at the correct time.


I recall one of the earlier space shuttle launches delayed for similiar reasons - a redundant set of onboard computers that didn't quite catch what each other was saying. One textbook I have on digital circuit design says 'there are three important concerns - timing, timing and timing'. :-)

As for Hanford :

Quote:

H1 is back!
Picking up where Jeff left off, in consultation with Gerardo I went out to the floor and checked on the POB photodiode cabling at the racks. Found that the idiot light LED for the BSCPD_01 cable (POB) was out. Did a quick check in the back of the rack for blown fuses, and found none. Called Richard and he pointed me to the problem right away - the POB2 photodiode was 'disabled' from the H1_LSC_PD_Interface.adl, an apparent oversight from the work earlier this afternoon. Enabled that, and H1 was back to locking in short order.

On the bright side, H1 is now back up after an extended period off this evening. Even better, the range is now running 17+ Mpc in segment #619, for over two hours now.


So it's great to see it running again ( I did sense some angst ). These IFO's are complex beasts indeed.

Now let's look at the lengths in this picture :

LA, LB, L1, L2, L3, L4 can all be changed individually by moving one or more mirrors. We'd call these independent length degrees of freedom. But while retaining the idea of independence, we could instead use combinations of these lengths :

Lccav = common cavity length = ( LA + LA ) / 2

Ldcav = differential cavity length = ( LA - LB ) / 2

Lcmic = common michelson length = ( L2 + L3 ) / 2

Ldmic = differential michelson length = ( L2 - L3 ) / 2

Lprc = power recycling cavity length = L1 + ( L2 + L3 ) / 2 = L1 + Lcmic

Lsrc = signal recycling cavity length = L4 + ( L2 + L3 ) / 2 = L4 + Lcmic

Don't worry too much about the factors of 2. The common lengths are a sum, the differential lengths are a difference.

We still have 6 degrees of freedom in length. In our minds we can think of any one of these varying, possibly while all the others are constant. But that's not to say the required mirror movements are simple to alter one of these composite lengths while keeping the others constant.

For instance if I wanted to change Lccav, while leaving Ldcav alone - I would alter LA and LB by the same amount and in the same sense. The sum of these two lengths would change but not their difference. So I could enact that by moving ETMA and ETMB both toward/away from the corner station where the beam splitter is. This will not change any of L1, L2, L3, or L4 either.

Ldcav is particularly important as the phase differences between the arms accumulate over many transits ( Fabry-Perot ). As far as GW reception goes this is where the money is.

Lccav is available for say keeping FP cavities in resonance while some tidal changes alter the relations between the end stations and the corner. Or some change in the laser frequency even.

Lcmic is not terribly interesting as the IFO doesn't really change character if this alters alone. You could do this by shifting ITMA, ETMA, ITMB and ETMB all together in the same sense and amount toward/way from the corner station. As we're keeping Ldmic constant then we don't introduce a shift in phase difference between the two arms.

What do you think of Ldmic?

Remember PRM is there to flick/keep the photons back into the gadget. It's position is adjusted to match phase relationships of light coming from the left, with those from it's right. In fact you could keep plain L1 as a degree : you'd leave the BS where it was ( or as determined by other ideas ) and alter where PRM sits. Ditto for L4 and the SRM. Remember SRM doesn't exist for Enhanced LIGO.

Indeed this entire discussion doesn't reveal the exact current operation ( which I don't know the setup and/or control code for ), but speaks of the general design flavour.

Remember the sidebands? These are frequencies of light from the laser, modulated by Pockel's cells, that are 'nearby' the base/carrier. One could adjust Lprc so that both the carrier and a sideband would be resonant. This is a similiar problem to finding least common multiples in integer mathematics. Take the numbers 3, 4 and 5 : you can find multiples of any of these separately, but is there a number which is a multiple of all three? Certainly! 60 for instance, and thus any multiple of 60 too. ( Bricklayers do this regularly by finding lengths that have convenient relationships in order to match the alignment of the bricks when separate courses meet. You don't have to, but it reduces the stuffing about with fractional bricks and aids the cohesion of the join ). All I want is the photons of different frequencies all coming back to the reflective surface of the PRM in the right phase.

You could imagine being some little bit of code whose job it is to read some photodiode level - say telling me how much light is leaking back out via the PRM - then deducing a direction and amount to alter the PRM's position - thus sending out control signals to the actuators ( say magnets & coils attached to the PRM ) - in order to restore the received PD level to some nominated 'ideal' setting. A cycle of read/calculate/adjust ..... ad infinitum. A feedback loop!

You could be yet another bit of code that holds the common cavity length at some preset - which another piece of code could decide the desired value of! I'd be fiddling with more mirror positions in this case, but the key point is that it's do-able without altering the other degrees. In fact as the IFO progresses from 'cold' to power up then to lock etc it is really the successive engagement ( I don't know the exact sequence ) and 'settling down' of feedback loops toward their respective 'preset goals', in order to finally achieve science data mode.

Enough for now! I'll read up on the more difficult alignment issue - the setting of the angles of the mirrors. The precision required is to do far better than simply hitting the barn door! :-) :-)

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

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6,580
Credit: 307,755,034
RAC: 187,595

Here's some further

Here's some further interesting points on timing from Livingston :

Quote:

ITMs, MCs restarted with new LOS code
We attempted to put the buffered communication, using doubles instead of floats, into the ASC to LOS communication links (ETMs, ITMs, RM, BS, MC2), but the load incurred put the CPU time at 58us. This is on the edge, so we decided to back off and make a smaller change.

The corner station optics are only reading the ASC data once every 8 cycles. Strangely, pitch was being read on the 2nd cycle and yaw on the 4th cycle. In terms of sensitivity to the ASC write time, reading only on the 1st cycle should be the best since the ASC writes near the end of its cycle (280us). We moved both reads to the 1st cycle and restarted l1dscl2 and 3 (ITMs and MC2) without problems. We should restart l1dscl1 (RM and BS) tomorrow.

The ETMs are a different story: they read the ASC correction signals every cycle. This makes them more likely to have problems, so we tried to make them only read on the 1st cycle, but we were unable to restart the FEs on the new code. We should test rebuilding with no code changes to be sure that the DSCs will start on a new build (from old code). Also TBD tomorrow.

Finally, the sensor signals for the corner station optics all go through the SOS FE (l1dscl0), which runs at 2kHz. These are read once every 8 cycles. We made no change to when they are read, but they could suffer from timing jitter in the SOS FE, which runs around 230us. These signals are pos, pitch, yaw, and side. Numbering cycles 0-7, they are read in cycles 2,3,4 and 5 (with ASC read in cycle 0). Jitter around 230us will probably cause yaw to glitch. We should consider repacking these if we can think of a good way to do it.

We will consider adding ASC and SOS generated cycle counters to the communication so that some error checking can be performed, if the load is not too great.


Meaning the choice of data representation ( double vs float ) led the CPU to be too close for comfort ( 58us vs 60us ) in it's realtime role. It has to complete the work before the result needs to be read! The other issue is that information used in correcting the IFO needs to be timely in all features. The timing jitter refers to some of the data items 'missing the bus' on some occasions.

Quote:

RFM delay
Thinking about this some more, the RFM delay from OMC to LSC is probably long enough to matter. Recall that the RFM works by handing messages around a fiber ring. Looking at the 5565 documentation, we find that the token hand-off time is 700ns. Since the fibers here at LLO go back and forth from the FEs to the by-pass switch, there is probably about 300ns of light travel time between FEs, so the total delay for each FE is close to 1us. There are 12 FEs on the ring, so roughly 12us is required to go around.

The OMC comes just after the LSC on the ring, so the time required for a change at the OMC to be reflected in the LSC memory is about 12us. That is long enough for writes in the odd OMC cycles to arrive after the LSC has started its next cycle. This plausibly explains why the OMC cannot write in odd cycles and have the LSC receive the data in time.


Recall that the RFM is reflective memory - a trick to make several data caches look like a single coherent one by rapid update of copies around a dedicated fibre optic ring. Allegedly this means the control loops are all reading off the same page ie. have a near simultaneous understanding of the instantaneous configuration of the IFO. The issue here is that the ring has one way of travel, clockwise say, and the LSC comes before OMC in that sequence. Hence data changes to RFM made by OMC take nearly a full circuit of the ring for LSC to know of that. It's like resetting a clock by only winding clockwise - if you want to change the time from 3 o'clock to 2 o'clock then you have to go right around through 12 rather than simply go back one hour. By the time LSC gets an update it's nearly/possibly old news.

This brings up a more general point or question : if you have photons tearing up and down the arms, at the speed of light obviously, then you can't beat them to the ETM's by some decision you make at the corner station. In this regard it seems the extra delay of communicating with the end stations vs the corner is important : 4km = 4000m ~ 12000feet ~ 12000 nanoseconds at c = 12us ( NB. the rule of thumb c = 1 foot per nanosecond ). This suggests the strategy that most, if possible, of the on-the-fly alterations should be decided upon and effected locally at the corner station ( basically among the PRM, SRM, BS and ITM's ). You can't get away from needing to tell the ETM's what to do, but I think the FP recycling cavity helps here as the storage time of the light is some ~100 arm transits. Hence if a gravity wave is passing through, of sufficiently long wavelength, we don't need to change the FP cavity length too quickly to stay resonant and ETM adjustments are thus less urgent.

I wouldn't mind seeing a timing diagram for this. It reminds me of Napoleon, and other generals of his era, who had to account for delays in receiving knowledge and the sending of commands. So when issuing some deployment order to counter some enemy move, you have to guess how the situation at some location had altered since the news ( that you just got ) had left that location. And then decide what would be the likely situation at that location when the troops arrived there at some future time based upon an order you are about to give!! Whew!! I reckon there were alot of conditional commands issued - if this do that, if not do the other ....

You may need to recall that the GW signal is deduced from the alterations that need to be made to keep the FP cavities resonant. DARM - Differential Arm Error - is the corrective signal applied to do this, and is effectively the science output of the IFO. This is a bit like using the deflections applied to a boat's rudder, in order to keep a straight course, as an indicator of the impact of waves on the vessel.

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

Dan G.
Dan G.
Joined: 10 Feb 07
Posts: 39
Credit: 97,382
RAC: 0

Hi Mike, can you point me to

Message 95182 in response to message 95181

Hi Mike, can you point me to where you're getting all this great info? I read the LIGO logs too and Weekly Reports, but I don't see all of this. A while back, I was reading a Caltech Ph.D. thesis about IFO lock technology in order to learn about LIGO optics. I'd love to learn about more resources.

Thanks!!

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6,580
Credit: 307,755,034
RAC: 187,595

RE: Hi Mike, can you point

Message 95183 in response to message 95182

Quote:

Hi Mike, can you point me to where you're getting all this great info? I read the LIGO logs too and Weekly Reports, but I don't see all of this. A while back, I was reading a Caltech Ph.D. thesis about IFO lock technology in order to learn about LIGO optics. I'd love to learn about more resources.

Thanks!!


It took a while for me to dig out stuff. I started here, and then found the best branches to particularly be this, that and the other. There's a fair amount at quite a range of levels and detail, especially this gold bearing seam so beware!! :-)

I also bought this book via Amazon. Peter Saulson writes well, but you'd need at least senior high school physics and/or university undergraduate level to get good value from it.

Also try this, which is where Bruce Allen currently is.

Enjoy! :-)

Cheers, Mike.

( edit ) This site is great for a quick ( or long! ) lookup on everything optics/laser/hardware. Especially for some of the less obvious terminology! You could start here for instance and wander off on some path ... :-)

( edit ) .... and please do speak up if you think I've got something wrong. It's a complex machine and it takes a good deal of synthesis to understand properly. I'm happy to be corrected as that only helps! :-)

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

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6,580
Credit: 307,755,034
RAC: 187,595

Now that I think about it I

Now that I think about it I might try to clear out a road block for you. In anticipation. I know it caused me a lot of angst. It's not a terrible problem, but it can be an unsettling issue when trying to get one's head around the whole GW thingy.

There are two equivalent viewpoints of gravitational waves. Some treatments aren't overt in which view is being used. The physically measurable results are identical either way. For this discussion let's ignore all those pesky sources of error and disturbance, and focus on GW's passing through only.

#1 Here I am sitting around using a 'usual' reference system. Say I'm at an IFO corner station. I've surveyed matters as precise as I can. When undisturbed my mirrors, dangling vertically from their suspending wires, aren't moving as I perceive them. A GW comes along and sets them in motion, with my laser system telling me that. My reference distances haven't altered, but the position of the mirrors with respect to that have.

#2 I choose to define my mirrors as 'freely falling', and also I am saying that my coordinate system is defined by such freely falling bodies. That means they suffer no non-gravitational forces. But actually they do, via those suspending wires. However the mirrors aren't restrained along the axis of motion that I will examine, in the horizontal plane. I consider that the mirrors are a fixed distance apart. A GW comes along and changes the meaning of a given separation though. So the mirrors haven't moved with respect to my markers, it's the space between the markers that has expanded/contracted! A bit obtuse perhaps. This is called the TT gauge view, for 'transverse traceless', a phrase based on some arcane language. 'Transverse' means the expansion/contraction is at right angles to the direction of motion of the wave - compare with longitudinal compression/rarefaction of air density along the line of propagation of a sound wave, say. I'm not very clear upon what 'traceless' implies, but I think it's something to do with how 'worldlines' are represented on diagrams.

The viewpoints are the same basically due to the Equivalence Principle! I can be in a non-free-falling frame - sitting in my chair getting supported by electromagnetic forces - and view the mirrors accelerating with respect to my unaltered reference frame. Or I can tag along with the mirrors and observe how their separation remains the same ( because I defined them to be that way ), but note that space is warped.

The absolutely key point is that light is indifferent to this. The phase evolution from mirror to mirror in either case is the same.

Cheers, Mike.

( edit ) Nah! A tensor is like a matrix that transforms between co-ordinate systems in 'good' ways - it preserves the 'character' of the problem, and we don't have to worry about being fooled by some inadvertent/special choice of reference frame. So the Earth has the certain geometric properties that we all love and know irrespective of how one sets up co-ordinates to describe it. The trace of a tensor is a way of combining it with itself, similiar to a vector dot product. You get a scalar - a single number. A vector dot product gives us the square of the length of the vector, which is an invariant across all choices of co-ordinate systems. That's what we want for physical system descriptions when we relate different viewpoints. A traceless tensor is one that has trace = zero. For our problem that relates to masses not accelerating with respect to the co-ordinate frame when a GW goes by .... so they don't move!!

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

Comment viewing options

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