Hannover Diary

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6534
Credit: 284700169
RAC: 113726
Topic 195856

Herewith a report upon the recent event at AEI/MPG in Hannover :

- it's a lovely 'university' city to visit, about 500,000 population centered upon many public parks and facilities with ample opportunity to walk and/or ride a bike virtually anywhere. Great trams. A delight to wander around.

- AEI is in the suburb of Herrenhausen as part of the university ( Nordstadt ) grounds. There is the Atlas cluster and optics labs on the grounds nearby. GEO is down the road about 30 mins drive near the small village of Ruthe.

- the event was held on Saturday and comprised a series of talks, plus tours of the cluster and labs however GEO wasn't available that day for a visit/tour due to being in science mode. Also a nice meal and free booze and chatting afterwards.

GEO : I did get a tour within a small group on the Friday during a one hour down time period for weekly maintenance. I thank the on-site GEO staff - especially Hartmut Grote who knows all things GEO having been with it since construction - for allowing this great opportunity. The longer/larger IFO's are not recording science data presently as they are upgrading, so it is important that GEO and Virgo remain 'ears up' as much as possible in that interregnum. Here is looking back up the access road next to the X-arm beam enclosure ( left side ) :

and you have to creep along below about 5km/hr to avoid bumping the IFO's 'lock'. Each arm is 600m long but there is 'folding' of the optical path ie. extra mirrors which effectively doubles that to 1.2 km. That's not all good news though as that folding can magnify effects under the general heading of 'displacement noise'. The corner station is in this low tech looking building, with the Y-arm going away to the right behind bushes n' stuff :

where you'll note a couple of obvious noise sources in addition to the adjacent worked farmland. There are several communications routes to AEI, this is to a geostationary satellite the other being a microwave link :

Here is a demonstrator for the seismic isolators being basically an inverted pendulum from which you hang a pendulum on a pendulum :

which have blade springs at the top :

which immediately suspend this mass :

and the second hangs from it. Here are actively controlled transducers which dampen the wobbles :

Inside the corner building lies a sealed and climate controlled area where the beams originate and recombine. This is deemed a clean area - we already had overshoes, but would have required further optics protective measures to go deeper in ( you can see my check shirt reflected in a nearside plexiglas type screen ) :

which I think I've correctly marked the beam orientation upon. IIRC inside this 'pot' is where the dark/anti-symmetric port readout is :

Here is the dark port photo-detector output during progression towards lock :

and I have no idea what the upper right hand date & time refers too ! AEI/GEO have developed and tested many parts present in the other IFO's like the laser for instance. Next time - the Atlas tour. :-)

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: 6534
Credit: 284700169
RAC: 113726

Hannover Diary

Atlas : A supercomputing cluster in a basement at AEI. Ever noticed that the good stuff is always 'downstairs' ? :-)

It's a large room with racks in rows :

you'll note the convergence of cables thus :

.... yes, it is Ethernet, including some hardware from a mob that went broke ( Woven Systems ) . This a 'trivially parallel' arrangement in the sense of quite separate work/jobs able to be independently set per node of which each is a quad core. There are currently 1680 nodes, each accessible on the spot via a 'crash cart' :

or more likely from the 'situation room' run by these two 'Masters of Disaster' aka Drs Henning Fehrmann and Carsten Aulbert who have lived and breathed this cluster for some years now :

where there was slightly too much red on the upper left of the 'threat board', but zero nodes down, no entries in the Hall Of Shame and of course the obligatory XKCD cartoon ..... :-)

Here's part of the ~ 3 - 4 Petabytes of storage, and in the cage is a robot that handles even way more of tape backup ( 2 copies of everything ) :

Now one can't forget infrastructure, so there's a battery UPS ( about 8 minutes worth for 'graceful' shutdown of the cores ) and a heat exchanger ( the cores automatically shut down quick slick if it fails )

The cooling is done by piping cold water to heat exchangers in every rack + blowing the air over it within the cabinets :

And I believe some expansion of the cluster is in the pipeline .... :-)

Next up is the optics lab.

Cheers, Mike.

( edit ) I've described the GPU component here.

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

astro-marwil
astro-marwil
Joined: 28 May 05
Posts: 511
Credit: 402260833
RAC: 1070936

Hallo Mike! Thanks for your

Hallo Mike!
Thanks for your reports and these impressive pictures. They show to me, what I have missed, as I couldn´t come as planed due to some havy malaise. I hope, you enjoyed your journey.

I would be pleased to hear something about the lectures halted.

Kind regards
Martin

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6534
Credit: 284700169
RAC: 113726

RE: Hallo Mike! Thanks for

Quote:
Hallo Mike!
Thanks for your reports and these impressive pictures. They show to me, what I have missed, as I couldn´t come as planed due to some havy malaise. I hope, you enjoyed your journey.


I did! I managed 6 hours sleep midway both legs, so that helped a great deal with the lag. Hannover itself was a hoot, and I also took the fast train each way between Frankfurt and Hannover. They have deer in the countryside like we have kangaroos .... :-)

Quote:
I would be pleased to hear something about the lectures halted.


Sure thing.

There were six lectures really, five on the day of the event and on Monday I got to sit in upon Ben Knispel's doctoral thesis presentation/defense ( which was a longer version of his Saturday presentation ).

Bruce kicked off Saturday with an overview of E@H in the last six years and the announcement of the new Fermi LAT work for E@H. How exciting! The satellite used to be called GLAST ( Gamma ray Large Area Space Telescope ) but was renamed in honor of Enrico Fermi. I've been reading up on the specs for this gadget which are quite impressive. You can't focus gamma's per se, so when one passes thru the detector - nearby a heavy metal nucleus - then it may convert to an electron/positron pair. It is this pair that is detected by a layered grid arrangement of plates backed by a calorimeter crystal. This way you get an energy and a direction to the sky source for each gamma photon. The surrounds of the detector have a trigger for cosmic rays ( gammas won't fire that up ) so serves to reject - 'anti-coincidence' - those sources, plus onboard electronics do some other sorting of candidates with the best being punted along the downlink. So that gives the data stream for our analysis. What are gamma ray pulsars? Well they are periodic in the way that the radio ones are, by definition, but one might assume the electrons that get ripped off the neutron star crust and follow the magnetic fields lines will be of higher energy and/or accelerated more to yield the synchrotron radiation that we see beaming toward us. I guess that suggests a more compact object?

Then there was Bikeman giving a great explanation of why it is that one can't just throw anything at a fast video card and expect a good result ie. can one massively parallelise? Maybe you'd like to come in on this thread HB ?

Michael Kramer came to talk to us about the exciting new prospects for the Square Kilometer Array - I cross my fingers for an Aussie siting of this - a ground based radio telescope array. The centre is a square kilometer but it also has 'arms' that sweep away from that for ? ~ 1000 km ? It is a coherent array meaning that we get phase information from each component ( ten meter dish ) so can effectively 'steer' the beam. Not to forget that many targets can be observed simultaneously by sub-arrays ! I will surely read up more on this, and I suspect there could well be a place for E@H in this scheme one day. :-)

Jonathan Leong talked of the design aspects of GW IFO's. He amusingly demonstrated how a high frequency oscillation will be barely absorbed by a lower fundamental frequency pendulum, and the active damping of a laser beam using feedback control to keep pointing steady. Recall that for the ET that will be a mirror 10km away! I think I understand 'light squeezing' now : one can arrange matters to 'park' the laser noise in one 'quadrature'. One way to represent Fourier series is with sines and cosines. Essentially one can go 'below' the uncertainty principle limit for noise in the sine component at the expense of blowing it out in the cosine part. Or put another way, on the Argand plane using a complex representation, one can think of noise compression along say the x-axis dimension going under Heisenberg but in compensation increasing in the perpendicular y-axis direction. In any case one just works with the less noisier part of the cycle for IFO purposes, pretending that you don't look at the noisy bit. I think .... :-) :-)

Now Ben Knispel's PhD presentation was a real treat, a convenient coincidence as regards timing the event. His research using E@H led to those two radio pulsar detections last year ( I was pleased to meet the Colvin's ) and naturally E@H enabled his data processing. I reckon this is a solid example of credible science done with the assistance of volunteer computing ie. the two-way street that 'citizen science' represents. He gave an overview of radio pulsars and how they come to be in terms of stellar evolution, the fraction that have the right geometry for us to perceive, and how to scan the sky for candidate objects. One key point is the Dispersion Modulus which is code/proxy for the length of the column of space between us and the source - containing free electrons that alter the pulse with the effect of higher frequencies arriving before lower ones. So one has to take guesses at the DM - our work units have that as a search parameter - in the hope that by correcting for that ( de-dispersing ) one can obtain a 'purer note' that can be identified as a pulsar waveform. There is a lot of 'hack work' in this type of research, checks and checks and checks. He made a good fist of the presentation, passed for sure, and had this great decorated 'mortar board' hat for us to inspect at the post-defense booze-up. :-0

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

Comment viewing options

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