Let me guess.... If I came to you in your chosen profession and said, "It hurts when I do this!", would you mime your best Groucho Marx cigar and say, "Well... don't DO that!" ;)
I like Groucho's 'This is so simple a six year old could understand it. Quick, go out and find me a six year old!' :-)
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
...I think it safe to say it's the same general aim - find GW signals, or upper bounds thereof, on continuous wave sources ( basically those with a persistent enough regularity in their emission to be detectable as such ).
Cheers, Mike.
Thanks! I guess a question would be is this new analysis considered more sensitive in digging the signals out of the data, or perhaps it is just another way of approaching the problem, or perhaps it is analyzing data that has not been analyzed previously?
The latter. The S5 LSC Science run lasted about two and a half years, when we started S5R2/3 only the data of the first year was available. S5R4 is looking at the rest of the data from S5, where the sensitivity of the detectors was higher than in the first year. Also we got more usable data out of the second year, which improves the sensitivity of the analysis. That, however, also means that for the workunits the data volume has increased and the amount of computational work necessary to process them, too.
BM
To demonstrate this, let's estimate how this affects runtime:
In S5R3, the WUs have used a grid of 5 spindown values (the rate at which the rotation of a pulsar changes over time, causing a shift in frequency of the assumed gravitational wave it is producing.
In the S5R3 WUs the following spindowns were used in the search templates (in Hz/s ):
So this should now explain the observed runtime behavior of the S5R4 units in comparison to the S5R3 units:
The total observation time of the new units is greater by a factor of ca 1.44 AFAIK.
The number of spindown grid points increased by a factor of 9/5 = 1.8
Runtime should grow roughly linear with number of skypoints, number of spindown grid points, and observation time. So per skypoint, the S5R4 units should take about 1.44 x 1.8 ~ 2.6 times longer, as there is 2.6 times the "science" done per skypoint, so to speak.
The S5R3 units typically contained 1200 skypoints, the S5R3 units have a varying number of skypoints, a typical one can have around 810 points. For the total workunit, the time it takes to process a S5R4 one should therefore roughly take
2.6 x 810 / 1200 ~ 1.75
times longer to process with comparable apps (s5r3 power apps and S5r4 stock apps are comparable).
This estimation doesn't match perfectly what I'm seeing with my hosts, but is reasonably close.
The reason I'm writing all this is to demonstrate that the longer runtime of the S5R4 units isn't demonstrating anything negative in terms of efficiency: There's just more science done per workunit.
To demonstrate this, let's estimate how this affects runtime: ... terms of efficiency: There's just more science done per workunit.
Well spotted. Those 9 spin down values cover much the same magnitudes as the previous 5. However if you line them up ( going from bigger to smaller down the page ) :
[pre] R3 R4
---------------------------------------
-3.171e-09
-2.7825e-09
-2.3941e-09
-2.0056e-09
-1.6172e-09
-1.2287e-09
-1.5855e-09
-1.197e-09
-8.4029e-10
-8.0859e-10
-4.5185e-10
-4.2015e-10
-6.3397e-11
-3.1699e-11 [/pre]
Looks like someone has raised both the upper and lower range bounds by near as damn all to a factor of 2. The ratio of top/bottom ( either column ) is real close to 50. The latter looks like an 'octave' - an additive ( not multiplicative ) one - with the common step of 3.885e-10 ( again either column ) ..... hmmm so we have the same absolute gapping within the range but up by factor of 2. The 3.885e-10 is likely to be related to phase space coverage, ie. trying to NOT miss a signal. The overall range change ( higher and wider ) is probably related to some calculation/suspicion on the expected binary pair behaviours.
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
The overall range change ( higher and wider ) is probably related to some calculation/suspicion on the expected binary pair behaviours.
Cheers, Mike.
Honestly I dunno .. but AFAIK, E@H is not looking for binary pulsars, only "solo" ones. Binaries will complicate the parameter space even further. But IIRC, the upper and lower bounds on the spindown parameter grid are indeed mainly motivated by astrophysical reasonings: It's less likely to find a strong gravitational wave from a pulsar that has a really tiny rate of spindown (in relation to its spinning frequency) , because a GW would drain the pulsar's energy which is not consistent with a very low spindown. OTOH, pulsars with an "excessively" high spindown rate must be very young (you can't sustain a high rate of spindown for a very long time without stopping to spin, obviously :-) ). But young pulsars are probably still surrounded by visible supernova remnants which we would have detected by other means, and E@H is more targeted towards yet undiscovered pulsars AFAIK.
The number of intermediate values between the upper and lower bounds of the search range that we want to try is, IIRC, mainly driven by optimizing the sensitivity of the search.
Honestly I dunno .. Maybe a physicist can better explain all this.
My guess only, as you'd think binary systems would cause more bleed of rotational energy of either partner? Complex. Anyhow, this is like trying to listen for a bird call before you've ever heard a sample of it. A physicist's view would be terrific. :-)
Quote:
A mate of mine who is a birdwatcher of some years tells me that in the forest you usually hear them first and then pick them visually based upon that directional clue. Except that some birds have calls, the characteristics of which in a forest's sound altering environment, are sufficiently diffused/echoed/faded to confound deduction of direction. To human ears anyway, and more likely to some significant predators/competitors. And then there's the mimics ......
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
To continue this very interesting discussion, I understand that SETI@home has just started the Astropulse project which, among other purposes, is supposed to assist in detecting Pulsars. Furthermore, it's just been announced that the NASA Fermi Gamma-ray Space Telescope has discovered a number of Pulsars. Are we not duplicating efforts here?
To continue this very interesting discussion, I understand that SETI@home has just started the Astropulse project which, among other purposes, is supposed to assist in detecting Pulsars. Furthermore, it's just been announced that the NASA Fermi Gamma-ray Space Telescope has discovered a number of Pulsars. Are we not duplicating efforts here?
Astropulse is looking for short pulses in the EM spectrum as a means to detect extraterrestrial civilizations but may find pulsars while doing so.
As to gravitaional wave research, there are (at least) two sorts of searches for continuous GW, one is by looking at known astrophysical bodies like pulsars and check them for GW emissions. This kind of search will benefit from any other search that will discover new pulsars (as a by-product or by intention).
Einstein@Home engages a different kind of search, currently it performs an all-sky scan for GWs, also covering yet undiscovered pulsars. So any project that discovers new pulsars will not make E@H any easier, but will also not make it less useful.
I couldn't agree more! Only sometimes I wish I had a canary to take with me down that mine so I'd know when the discussion has reached a level of technical difficulty that will cause a brain aneurysm. ;-P
RE: Let me guess.... If I
)
I like Groucho's 'This is so simple a six year old could understand it. Quick, go out and find me a six year old!' :-)
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
RE: RE: RE: ...I think
)
To demonstrate this, let's estimate how this affects runtime:
In S5R3, the WUs have used a grid of 5 spindown values (the rate at which the rotation of a pulsar changes over time, causing a shift in frequency of the assumed gravitational wave it is producing.
In the S5R3 WUs the following spindowns were used in the search templates (in Hz/s ):
-1.5855e-09
-1.197e-09
-8.0859e-10
-4.2015e-10
-3.1699e-11
==> 5 different values
In S5R4 , there are now 9 of them :
-6.3397e-11
-4.5185e-10
-8.4029e-10
-3.171e-09
-2.0056e-09
-2.3941e-09
-2.7825e-09
-1.2287e-09
-1.6172e-09
So this should now explain the observed runtime behavior of the S5R4 units in comparison to the S5R3 units:
The total observation time of the new units is greater by a factor of ca 1.44 AFAIK.
The number of spindown grid points increased by a factor of 9/5 = 1.8
Runtime should grow roughly linear with number of skypoints, number of spindown grid points, and observation time. So per skypoint, the S5R4 units should take about 1.44 x 1.8 ~ 2.6 times longer, as there is 2.6 times the "science" done per skypoint, so to speak.
The S5R3 units typically contained 1200 skypoints, the S5R3 units have a varying number of skypoints, a typical one can have around 810 points. For the total workunit, the time it takes to process a S5R4 one should therefore roughly take
2.6 x 810 / 1200 ~ 1.75
times longer to process with comparable apps (s5r3 power apps and S5r4 stock apps are comparable).
This estimation doesn't match perfectly what I'm seeing with my hosts, but is reasonably close.
The reason I'm writing all this is to demonstrate that the longer runtime of the S5R4 units isn't demonstrating anything negative in terms of efficiency: There's just more science done per workunit.
CU
Bikeman
RE: To demonstrate this,
)
Well spotted. Those 9 spin down values cover much the same magnitudes as the previous 5. However if you line them up ( going from bigger to smaller down the page ) :
[pre] R3 R4
---------------------------------------
-3.171e-09
-2.7825e-09
-2.3941e-09
-2.0056e-09
-1.6172e-09
-1.2287e-09
-1.5855e-09
-1.197e-09
-8.4029e-10
-8.0859e-10
-4.5185e-10
-4.2015e-10
-6.3397e-11
-3.1699e-11 [/pre]
Looks like someone has raised both the upper and lower range bounds by near as damn all to a factor of 2. The ratio of top/bottom ( either column ) is real close to 50. The latter looks like an 'octave' - an additive ( not multiplicative ) one - with the common step of 3.885e-10 ( again either column ) ..... hmmm so we have the same absolute gapping within the range but up by factor of 2. The 3.885e-10 is likely to be related to phase space coverage, ie. trying to NOT miss a signal. The overall range change ( higher and wider ) is probably related to some calculation/suspicion on the expected binary pair behaviours.
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
RE: The overall range
)
Honestly I dunno .. but AFAIK, E@H is not looking for binary pulsars, only "solo" ones. Binaries will complicate the parameter space even further. But IIRC, the upper and lower bounds on the spindown parameter grid are indeed mainly motivated by astrophysical reasonings: It's less likely to find a strong gravitational wave from a pulsar that has a really tiny rate of spindown (in relation to its spinning frequency) , because a GW would drain the pulsar's energy which is not consistent with a very low spindown. OTOH, pulsars with an "excessively" high spindown rate must be very young (you can't sustain a high rate of spindown for a very long time without stopping to spin, obviously :-) ). But young pulsars are probably still surrounded by visible supernova remnants which we would have detected by other means, and E@H is more targeted towards yet undiscovered pulsars AFAIK.
The number of intermediate values between the upper and lower bounds of the search range that we want to try is, IIRC, mainly driven by optimizing the sensitivity of the search.
Maybe a physicist can better explain all this...
CU
Bikeman
RE: Honestly I dunno ..
)
My guess only, as you'd think binary systems would cause more bleed of rotational energy of either partner? Complex. Anyhow, this is like trying to listen for a bird call before you've ever heard a sample of it. A physicist's view would be terrific. :-)
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
RE: A physicist's view
)
Point Reinhard to this thread or ask him via eMail. He did the major design of the S5R4 search and is pretty good in explaining these things.
If he hasn't got the time, contact Oliver Dreissigacker at AEI. He won't know from the top of his head but he will find out and explain.
And Bikeman is right, Einstein@home is not looking for binary systems, but just isolated spinning masses.
BM
BM
To continue this very
)
To continue this very interesting discussion, I understand that SETI@home has just started the Astropulse project which, among other purposes, is supposed to assist in detecting Pulsars. Furthermore, it's just been announced that the NASA Fermi Gamma-ray Space Telescope has discovered a number of Pulsars. Are we not duplicating efforts here?
Warped
RE: To continue this very
)
Not really (also see this discussion).
Astropulse is looking for short pulses in the EM spectrum as a means to detect extraterrestrial civilizations but may find pulsars while doing so.
As to gravitaional wave research, there are (at least) two sorts of searches for continuous GW, one is by looking at known astrophysical bodies like pulsars and check them for GW emissions. This kind of search will benefit from any other search that will discover new pulsars (as a by-product or by intention).
Einstein@Home engages a different kind of search, currently it performs an all-sky scan for GWs, also covering yet undiscovered pulsars. So any project that discovers new pulsars will not make E@H any easier, but will also not make it less useful.
CU
Bikeman
Thanks Bikeman. You remain
)
Thanks Bikeman.
You remain a mine of information.
Regards.
Warped
RE: Thanks Bikeman. You
)
I couldn't agree more! Only sometimes I wish I had a canary to take with me down that mine so I'd know when the discussion has reached a level of technical difficulty that will cause a brain aneurysm. ;-P