Detector Watch 2

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6,018
Credit: 96,592,860
RAC: 154,024
Topic 192002

This is a continuation of the previous thread.

It's time for a talk about time. :-)

Einstein probably said it all when saying "Time is that which is measured by clocks". Here one can get into alot of deep-ish water looking for a firmer definition - that is not a bad pursuit of itself, but it doesn't advance science much in practice. So Einstein tied the concept of time with it's measurement, pretty much like tying the premise of a 'rigid body' ruler to distance. I guess we all have an intuitive idea of time which is hard to capture precisely and consistently, so the science way is not to worry too much about that aspect, but to just create a working definition which covers our main sensibilities and work from there. So if we have a clock design that a priori ( in advance ) seems to be likely to produce a 'regular' signal then we go with that.

Any measurement of some clock's accuracy is only going to be made with reference to another clock, so like other fundamental quantities it is tied to some agreed standard. When I last looked a second was defined to be a certain number of oscillations of a light signal emitted during the transition of electrons from one atomic state to another ( Caesium I think ). There are a number of variants upon this technique - the short answer is that time ( as an international standard ) is based upon averaging from several clocks of that general type - safety in redundancy. That's not any better, or worse, than defining distance to be based upon a dead English king's stride ( Imperial Foot ) or a fraction of the circumference of the Earth ( the metre ).

However for general use, and LIGO in particular, a thingy called GPS time is used. As you probably know, the Global Positioning System relies on very accurate timing. A useful rule of thumb here is that light travels at about one foot per nanosecond - yes that's the stride of a dead English king in one billionth of a second. Nothing at all has been found to travel faster than that. So if you are, say, 60 feet from your neighbour then you are also 60 nanoseconds from them too. This introduces the tricky concept of when do things happen - when I record them or when my neighbour does? For everyday stuff who cares, eh?

Not so for many science purposes however. In general, separated observers ( people with recording instruments ) will disagree about whether events A and B ( stuff that happens ) occurred simultaneously, A before B, B before A, and how much by. Following the detail of such analyses leads into the formulation of the Special Theory of Relativity which is not the real path here. But I will talk of how one might in practice synchronise separated clocks.

A first attempt is to create two identical clocks - same design, materials etc - and tune them in to each other in some way so as they tick together. That is every advance of each clock's counter ( whatever it is ) is never different between the two when they are next to each other. Now leave one ( call it clock A ) at Site A and move the other ( called clock B ) to Site B. So they should be synchronised now, OK? Well not quite ..... as it leaves the possibility that some aspect of transport may have affected say clock B's function. I don't mean it got bumped in transit ( although that may have happened ), I refer to something more fundamental - the nature of the passage of time itself during the trip.

So to eliminate that possibility in our process, we try to signal between the clocks once they are set in their respective positions in order to then effect synchrony. Roughly speaking it goes like this:

- at A fire off a light signal to B: note down the moment on clock A that you did that ( say 12:00 PM exactly ).
- upon arrival at B: immediately respond with a signal back to A which also encodes the time showing on clock B when A's signal did arrive ( say 12:03 PM exactly ).
- when B's signal comes back to A: note the moment that occured ( say 12:04 PM exactly ) and discover whatever that encoded time was at B.

So then, at Site A, I'd have on my timesheet:

Event #1 - Signal left A: 12:00
Event #2 - Signal arrived at B: 12:03
Event #3 - Signal back at A: 12:04

NB. We can factor delays in response if you like, but it doesn't change the heart of the method. Also we can only record events at this stage using clocks placed at the position of those respective events - because we haven't yet achieved/decided how to compare/synchronise separated occurences yet!

So you will look over that sheet and deduce that our separated clocks are saying that it took 3 minutes for a signal to go one way but only 1 minute on the return leg. So inevitably we must conclude that either:

- there is some anisotropy ( directional dependence ) to our signalling

OR

- we need to adjust one or both clocks.

The solution preferred is to signal site B, again, and ask that clock B be wound back by one minute. This is because we conclude that the round trip time is 4 minutes ( as measured by clock A alone ), then by discarding anisotropy ( to date there is no evidence for it ), we assign 2 minutes per separate leg & equal each way. As clock B said 12:03, not 12:02, then that hints at the correct adjustment to it.

You can adjust this method for any other clock readings in similiar fashion. If that is then done we would consider, or define, that clocks A and B are synchronised. If you worry about some 'drift' in agreement over time, then repeat the above whenever you like and re-adjust, possibly fiddle with some 'rate' knob on the clocks etc... interpolate between readings .... and so on.

One characteristic of this approach is that if we have a third clock C, at a site C, which is twice as far from A as B is - and in the same direction ( ie. A, B, and C in an equally spaced line ) - then if we synchronise A and B, followed by B and C, then A and C must also be synchronised ( even if we never do the messenging! ). Thus this can give rise to a consistent, anisotropic, grid of clocks to work from. It will be consistent with theoretical aspects of time as used in Special/General Relativity. ( I'm talking about 'normal' space here, where a clock hasn't fallen into a black hole etc ).

Now you may be asking what the heck does this have to do with LIGO and GPS? In effect GPS is like the old 'speaking clock' on the phone systems - "at the third stroke it will be ...." - continuously broadcasting timing information via satellite and precision recievers. It plays the role of a clock saying 'I am here at position P and I say it is time X'. The GPS in turn relies, or calibrates, against the atomic clock standards mentioned above. Each LIGO installation distributes timing information within and is slaved to the GPS time channel. The data 'pipelines' regularly stamp the data 'frames' accordingly. This is absolutely crucial for any later deduction of the direction of a gravity wave's arrival at Earth.

I think the acceptable inaccuracy is no more than about 10 nanoseconds ( ten feet! ) across LIGO. The light flight time from Hanford to Livingstone is about 10 milliseconds ( ie. 10,000,000 feet or approximately 1800 miles ) on a direct line through the Earth ( which is curved after all ). I don't however know what angular resolution this yields for pinpointing sources in the celestial sky.

Cheers, Mike.

( edit ) Changed gif frame-rate from 14 to 8 ( per second ).

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

Chipper Q
Chipper Q
Joined: 20 Feb 05
Posts: 1,540
Credit: 708,571
RAC: 0

Detector Watch 2

Thanks for the proper perspective on timing, Mike! It looks like a major chore just doing the time-keeping; come to think of it, with correct time-stamps on the data frames for generating the E@H work units, it means that by virtue of the software we're running, all of us crunchers are helping out with the time-keeping and synchronization of the data that's a part of making an ultimate 'Figure of Merit' (FOM): a GW signal matching an inspiral template was/wasn't detected at location x and time t. Go BOINCers!!

Speaking of FOMs, I checked FOM1 from Hanford (Mike posted a sample of one of these here) and the State Vector (sample and good explanation of one of these here) showed that there was ~7 hours of time with both Livingston and Hanford in Science Mode, or ~7 hours in 'triple coincidence' (T0 was 29/10, 04:09:51 UTC). At one point, all 3 IFOs dropped out of SM, but I think for different reasons. The weather, as in the wind, was causing problems for the 2 interferometers at Hanford:

Quote:
Started 2 hours ago, died down for a little while, and is now reaching 30+mph at all stations and 40+ at End Y.


Also the End Test Mass in the 2-kilometer IFO, at the Y-end, known in the logs as ETMy, caused another drop in the Inspiral Range to 2.2 Mpc. Looks like it might be an electrical problem. Troubleshooting an intermittent electrical problem is very difficult. Many of us probably are familiar with the same kind of malfunction in our cars! The level of diligence and professionalism shown in the efforts to identify ETMy's problem inspires confidence that one of the most sensitive instruments in the world will continue to improve in the quality and quantity of data it gathers. "Have data, will crunch!":) There are quite a few things it could be (even including friction in the suspension), and it will be interesting to see what develops with ETMy.

S5 run, daily locked statistics for 28/10 at Hanford:
H1 was Duty 92.2%
H2 was Duty 71.4%
Not too shabby!

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6,018
Credit: 96,592,860
RAC: 154,024

I forgot to add that GPS time

I forgot to add that GPS time is based on counting seconds since GPS zero - which is Sunday January the 6th, 1980, at 00:00:00 UTC. There is a separate lookup table for inserting leap seconds that have accrued since, so that such corrections as applied to data frames depend on when those frames were produced. If you look at this bit:

of this plot:

then that is the GPS seconds count at the moment the plot was produced. You will see a similiar indicator in other places.

Note: I have seen proposals for alternative timing systems, say with atomic clocks and optical fibre distribution - for an in-house LIGO reference to reduce exposure to GPS errors which would seriously compromise GW analysis.

Each occasion they have a H2 error/issue it allows for further analysis of the problem(s) at ETMY. Would be nice without it, of course, but it certainly does exercise the LIGO crew's problem solving and team abilities - not to mention giving us bystanders a glimpse of the complexity of the interferometer (mal)functions! :-)

Cheers, Mike.

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

Chipper Q
Chipper Q
Joined: 20 Feb 05
Posts: 1,540
Credit: 708,571
RAC: 0

RE: I forgot to add that

Quote:
I forgot to add that GPS time is based on counting seconds


I noticed that the total time for the data segments tallied in seconds, as well. Certainly much easier working with intervals of time that way, rather than the MM/DD/YY hh:mm:ss format!

Looks like another pretty good day for data gathering - from Hanford's Owl Summary:

Quote:

Good triple coincidence tonight.

In the last 16 hours, H1 had just one science drop from an AS_Q trigger coinicident with a passing tandem trailer truck. H1's range has stayed close to 15 MPc tonight.

H2 has also been up in science mode for the last 16 hours with its range close to 7 Mpc.

Seismic in the last hour has been affecting both the ranges.


So EMTY in Hanford's 2K IFO has been working okay (at least for 16 hours)...

Quote:
Each occasion they have a H2 error/issue it allows for further analysis of the problem(s) at ETMY. Would be nice without it, of course, but it certainly does exercise the LIGO crew's problem solving and team abilities - not to mention giving us bystanders a glimpse of the complexity of the interferometer (mal)functions! :-)


...but I just noticed that Livingston has also been experiencing similar (AFAIK) behavior with their ETMY as well:

Quote:
In reading yesterday's elog , see that the ETMY bias
module has been bypassed and the output of the DAC is being fed to the
bias terminals of the coil driver chassis. This may be the cause of the excess noise in 50 to 100 Hz band. The bias module not only amplifies the DAC but also filters the high frequency noise in the DAC output which is very high. Strongly recommend that you replace the bias module with one that has been tested.


There was also this entry in Livingston's log:

Quote:

Excess noise between 40-120Hz led to a swap of the ETMY bias module this morning.

After the swap, the noise was unchanged. The noise was unaffected by changes to
the input power.

The excess noise is the same as that observed on Oct 13, 14, and 15. It is highly
nonstationary, but some noise is always present.

The microseism is elevated (well above the 50th percentile), but it's not the
cause of the noise. Good sensitivity and stability were achieved on a week ago on
Oct23, despite similar ground motion.

We've had intermittent newtork-related problems this morning. The upscript and
downscript crashed several times (killing locks in the process); these problems
seem to be fixed after the operator's workstation was rebooted. DMT has been very
slow: the trends are updating about once every five minutes.


Hmm... how soon before the upgrade from initial to advanced LIGO suspensions?

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6,018
Credit: 96,592,860
RAC: 154,024

As mentioned in the Hanford

As mentioned in the Hanford log this is indeed a 'nice example of seismic event and lockloss on H1':

The lower blue plot the spike is the ground shaking. The upper red plot is the light emitted back to the beam splitter from the Fabry Perot ( FP ) resonance section of the interferometer ( IFO ), that is the portion between the input ( ITMY ) and end ( ETMY ) test masses.

If the IFO is in lock then the light between the two test masses will bounce back and forth many, many times ( on average ) before returning to the corner station ( LVEA ) for comparison with light from the other arm ( see here for IFO layout ). This situation of resonance requires the positioning of the two masses/mirrors to within a fraction of the wavelength of the light used! This mechanism is where LIGO obtains it's sensitivity from.

If a seismic or other disturbance defeats the various isolation mechanisms surrounding the apparatus then resonance no longer occurs ( specifically the phases of the photons arriving to the mirror from within the FP cavity no longer sum up with each other in synchrony, but subtract ... )

Don't worry that I've showed the seismic record from the mid station of the X arm - it showed the clearest signal plot. The whole IFO is affected when the ground shakes and in this instance the Y arm dropped out. Clearly it only needs one arm to go off for useful operation to be lost. Such is the nature of the beast! :-)

Hanford has lost science mode time due to disturbances around the facility - various works under way.

Livingstone is similiarly affected, I don't know the detail but the seismic plots are busy at most frequancies, up until later in the shift.

Cheers Mike.

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

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6,018
Credit: 96,592,860
RAC: 154,024

Having talked of time, now we

Having talked of time, now we move to frequency. This seemingly mundane concept is very central to the LIGO project.
Strictly speaking the frequency of 'something' is only interesting if the 'something' repeats. If you have someone knock on your front door, once, and never again then the frequency of that event is zero. Not alot to talk about really. If however they return to repeat the act you can now sensibly talk of frequency. Generically, it is the ratio of two numbers:

frequency_of ( something ) = [number of times ( something ) happens in stated time interval] / [stated time interval]

By convention we usually use one ( 1 ) second as the time interval. So if something repeats once every one second, we call/define that to be one ( 1 ) Hertz. This is the fundamental frequency unit named in honor of Heinrich Hertz, the nineteenth century guy who ( amongst other achievements ) discovered the propagation of radio waves across the 'air', without wires, from a spark generator to a separate circuit.

A related idea is the period of something. Again if something happens only once, there's not alot to discuss. But if it recurs then the gap/interval/difference in time between one something and the next something is designated as the period. May as well measure it in seconds - you don't have to, but it's then easier to convert to Hertz if you need to by :

frequency = 1 / period

thus if something has a period of half a second, then it will happen twice in a second. If something has a frequency of 0.1 Hz ( one tenth of a Hertz ) then it recurs once every 10 seconds. For instance, if mention is made of 'millisecond pulsars' that is actually implying some signal has been received at about 1000Hz. One can apply the usual multiplier prefixes - Kilo ( 1,000 ), Mega ( 1,000,000 ) and Giga ( 1,000,000,000 ) etc - or divisor prefixes - Milli ( thousandth = 0.001 ), Micro ( millionth = 0.000001 ) and Nano ( billionth = 0.000000001 ). Get used to the idea of 'flipping' ( taking the reciprocal of ) to go from one to the other.

( There is some distinction to be made between discrete variables and continous ones, but I'll gloss over on that ... )

Check out these patterns ( plot Alpha ) :

they depict sine waves of differing frequencies, which respectively repeat one ( A ), two ( B ) and three ( C ) times per second. Hence they have periods, respectively, of one second ( A ), half a second ( B ) and one-third of a second ( C ).
I could add them all together and get ( plot Beta ):

Another way of examining this cyclic behaviour is to list/plot the frequencies at which these things recur, rather than having to count cycles and calculate ( plot Gamma ):

I've deliberately not specified the vertical axes in either case - it's not really important here - but it reflects the amplitude/strength of a something/signal in Alpha and Beta and it's power/density in Gamma.

We arrive at a marvellous thing in mathematics, the Fourier transform. This is named after the French bloke who invented it. It's a procedure ( and it's reverse ) for swapping between the Beta representation and the Gamma. Think of it like asking the question ( of any old signal ) - what sine waves, and how much of each, would I add together to get the observed waveform?

Funnily enough, we ( nearly ) all do this every day without realising it. Our ears pick up the energy of molecules hitting our eardrums and a special organ ( the cochlea ) breaks it down into frequencies for transmission to the brain. There are cells in the visual circuitry of our brains that selectively fire up if presented with certain ( spatial ) frequencies in the visual field. Small receptors, say at our feet, will distinguish between different modes of vibration of our skin. Etc...

Here's an example of the H1 and H2 interferometer output signals converted via Fourier transform:

You can see a number of peaks in this 'spectrum' which indicate some dominant frequencies in the output. Some of this is not necessarily desirable, they may represent non-astronomical sources.

Cheers Mike.

( edit ) My gosh, I almost forgot to mention! The frequency of gravitational waves significantly overlaps with human hearing range. So this chap has prepared sound files so that you can listen to the gravity waves as if our ears were LIGO detectors! I like the inspirals best.... :-)

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

Mahray
Mahray
Joined: 11 Nov 04
Posts: 43
Credit: 783,499
RAC: 0

Mike Thanks a lot for

Mike

Thanks a lot for putting this stuff together, it makes things much more interesting. Keep it up! (Please!!)

Chipper Q
Chipper Q
Joined: 20 Feb 05
Posts: 1,540
Credit: 708,571
RAC: 0

Cooler temperatures outside

Cooler temperatures outside at Hanford led to turning on a couple heaters in different areas (lab/control room, and also in the LVEA). This may cause subtle changes in the alignments, and of course, special attention is focused on the OPLEVs.

There was also a problem with FB1 (Frame Broadcaster), and this meant that the usual FOMs weren't available. I think this is related to a file system ('LDAS') that contains the scripts that are run to collect sensor data and then perform the necessary calculations for generating an FOM, e.g., millivoltage readings of a sensor are converted using calibration settings (set in the script) to produce a strain value, or the gain in a feedback circuit is converted by the script into millimeters of beam-offset from the center of an optic.

So various scripts are run automatically to produce the 'RoboMon Robo SciFom' elog entries, and I'm guessing that there's a library of additional scripts that can be used for special cases, and that new scripts (as with the new DARM_ERR plot) are written as the situation warrants. I also noticed that a special template was written to help investigate and identify which close seismic event is responsible for a loss of lock in the IFO. The plots Mike posted for 'nice example of seismic event and lockloss on H1' (above) were generated using that template.

When FB1 was down, and the usual FOMs were unavailable, a plot of the Noise Budget was generated to help provide an indication of the inspiral range, among other things. From looking at that, it's easy to see why there are additional FOMs normally available, as the Noise Budget is quite crowded. And with Mike's excellent explanation on frequency, you can get a fair idea of how many different noise sources must be accounted for in order to arrive at the waveform that possibly indicates a signal from a gravitational wave source. The peaks Mike pointed out in the IFO output signals (that may be due to non-astronomical sources) have to be checked against all the sources of noise... here's what the Noise Budget plot (for H1) looks like:
[img]http://groups.msn.com/_Secure/0SADjAkoW3z04YGEiWeXKE5dIchMM4rpfQJOPt!g7oxe7Vvn*LENnsJmIac485P5Rj49I*9EW6GC6zM3SHqeTC*rj1ym2KkSMtGcmWsLZVW8vAAAAynJbAg/NoiseBudget.jpg?dc=4675595941427881084[/img]

Chipper Q
Chipper Q
Joined: 20 Feb 05
Posts: 1,540
Credit: 708,571
RAC: 0

They're trying to identify a

They're trying to identify a source of ground noise at Hanford. It started yesterday, and wasn't sever enough to prevent lock, but when the noise finally stopped, the inspiral range in both IFOs improved (H1 climbed to ~15 Mpc, and H2 to ~7 Mpc). From the State Vector it looks like lots of time in Science Mode for Hanford.

At Livingston there were the usual late night trains (at 3 different times last night), and there was a bit of glitching getting back into SM after one of the trains; the laser power had to be adjusted down at first. Also during the Owl Shift was mention of a Gamma Ray Burst trigger. (SWIFT caught 2 so far today, one at 01:00:31 UTC and one at 06:13:16 UTC.)

Chipper Q
Chipper Q
Joined: 20 Feb 05
Posts: 1,540
Credit: 708,571
RAC: 0

Found this entry in Hanford's

Found this entry in Hanford's Swing Shift Summary:

Quote:

Hands off shift, 2k continues with segment H2-1833 and the 4k continues with H1-2427. Inspiral range is at 14.5 Mpc for H1 and 6.8 Mpc for H2.

Received two GRBs, both of the IFOs were locked and in SM.

Livingston's doing well, all things considered. Their FOM1 is slightly different from Hanford's, and has a plot of anthropogenic noise. Quite obvious to see when a train goes by; I added a label and arrows to this plot:

The following Livingston elog entry is one of the most exciting ones I've seen yet (although only been watching the detectors regularly since Mike started the 1st DW thread)

Quote:

Coincident H1 / L1 spectral lines found in pulsar analysis
The PowerFlux pulsar search program (V. Dergachev) has been
run over the first 8 months of the S5 data, searching for
0-spindown pulsars from 50-1000 Hz. In the preprocessing
step, there is an on-the-fly line flagging, based on months
of integrated 30-minute SFT's of h(t).

This report lists stationary or quasi-stationary lines found to be
coincident within 10 mHz between H1 and L1, coincidence close enough
that they could mimic pulsars in regions of the sky near the
ecliptic poles. The number of coincident candidates grows very rapidly
as one lowers the SNR threshold from each IFO. Here we show results
for two thresholds which correspond (approximately) to 6 and 5 sigma.
There are two cases (marked by asterisks) where the lines wander enough
to show up as putative pulsar candidates at the end of the analysis,
(but which do not survive further examination).


There was a sizable list of coincident frequencies. If any of them match the location of a known pulsar, does this mean the LIGOs are detecting GWs?

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6,018
Credit: 96,592,860
RAC: 154,024

Now here's a tasty tidbit

Now here's a tasty tidbit from the Hanford log ( Thu Nov 2 21:37:49 2006 UTC ):

Quote:

Coincident H1 / L1 spectral lines found in pulsar analysis

The PowerFlux pulsar search program (V. Dergachev) has been
run over the first 8 months of the S5 data, searching for
0-spindown pulsars from 50-1000 Hz. In the preprocessing
step, there is an on-the-fly line flagging, based on months
of integrated 30-minute SFT's of h(t).

This report lists stationary or quasi-stationary lines found to be
coincident within 10 mHz between H1 and L1, coincidence close enough
that they could mimic pulsars in regions of the sky near the
ecliptic poles. The number of coincident candidates grows very rapidly
as one lowers the SNR threshold from each IFO. Here we show results
for two thresholds which correspond (approximately) to 6 and 5 sigma.
There are two cases (marked by asterisks) where the lines wander enough
to show up as putative pulsar candidates at the end of the analysis,
(but which do not survive further examination).

>6 sigma:

H1 L1
-- --
64.000 Hz 64.000 Hz (4th harmonic of 16 Hz)
85.803 Hz 85.800 Hz
100.000 Hz 100.000 Hz
112.000 Hz 112.000 Hz (7th harmonic of 16 Hz)
127.310 Hz * 127.304 Hz
128.000 Hz 128.000 Hz (8th harmonic of 16 Hz)
205.318 Hz * 205.318 Hz
231.576 Hz 231.584 Hz
236.714 Hz 231.723 Hz
424.577 Hz 424.587 Hz
455.098 Hz 455.089 Hz
459.358 Hz 459.363 Hz
514.437 Hz 514.434 Hz
544.669 Hz 544.666 Hz

.....


There's a much longer list for the 5 sigma group.
May I particularly emphasise the use of the words candidate(s), putative and mimic above. These are not detection claims. However this gives a feel/feedback on the processing. What is discussed here are signals received within a narrow band of frequency, recorded both by Hanford and Livingstone 4K LIGO's. They have appeared in the preliminary steps of analysis of searches for pulsar signals in the 50 - 1000Hz range, and could be consistent with pulsars emitting more or less directly above or below the plane of Earth's orbit.

- SNR means signal to noise ratio, roughly indicating how far the desired signal 'tree' grows above the 'forest' of other sources.

- SFT means 'Short Fourier Transform' performed on 30 minute segments of data in the initial 8 months of S5.

- h(t) is the ( presumed ) gravitational wave strain on spacetime.

- 'sigma' refers to a statistical parameter used to gauge the likehood of some outcome ( here a 'real signal' ) vs some base case ( here 'random noise'/'nothing'/'something else' ) - a higher number ( 6 vs 5 ) suggests greater confidence in a detection essentially.

Who knows, which ( if any ) of these lines will pass subsequent hurdles of proof! :-)

Cheers, Mike.

( edit ) Ha! Double hit there, Chipper! It certainly is a head turner.... :-)

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

Comment viewing options

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