Thanks Ben, I have my missing bits now. A real signal may still line up in the bins after some de-dispersion attempt, now in 8-bit dynamic range and also remembering that the pulsar emissions are fairly broad across the bands of interest anyway. So that power, added in quadrature, if periodic in the time series can then be caught by Fourier etc. A good statistic plus details from that goes toward new observations to confirm and characterise maybe a new discovery. Actually that makes the process and its proven successes even more impressive! :-) :-)
One other question if I may : what's the order of magnitude of a 'typical' dispersion delay ( or the difference b/w that for lowest frequency of interest c/w the highest ) compared to the sampling interval ? Or if you like : what's the general range of how many samples do we shift the data across to successfully get the 'right' emission spectrum ? Hopefully I have worded that right .... ie. I don't know how to convert a delta_DM value to a time delay.
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
One other question if I may : what's the order of magnitude of a 'typical' dispersion delay ( or the difference b/w that for lowest frequency of interest c/w the highest ) compared to the sampling interval ? Or if you like : what's the general range of how many samples do we shift the data across to successfully get the 'right' emission spectrum ? Hopefully I have worded that right .... ie. I don't know how to convert a delta_DM value to a time delay.
The dispersion delay between two radio frequencies f1 and f2 is
delta_t = 4149 secs * (f1^-2 - f2^-2) * DM,
where DM is measured in pc/cm^3 and f1 and f2 are measured in MHz. You can divide this by sampling time to get the dispersion time delay delta_t in samples. You can find this formula for example on this handy NRAO webpage. It's also equation (2.43) in the introduction of my PhD thesis.
For the PMPS data we have f1= 1230 MHz, f2= 1518 MHz, and for a DM of, say, 200 pc/cm^3 I get delta_t = 0.188 seconds, which corresponds to 753.5 of the 250-microsecond samples.
Hi all,
the following is to give you a bit more background on the new binary radio pulsar search "BRP6" aka "PMPS XT" is about. The data your GPUs will be analyzing are archival observations from the Parkes Telescope in Australia, from the very successful Parkes Multi-beam Pulsar Survey (PMPS).
Thank you much for the very helpful information regarding the new binary radio pulsar search.
As announced earlier the successor to the "S6Bucket Follow-up #1" (FU1) will be another follow-up on the candidates we got from the S6Bucket search, zooming even deeper into the top candidates from the "Follow-up #1" (FU2). Whereas the FU1 tasks consisted of 8 independent "atomic tasks" bundled together with an (initially rough) estimated runtime of 2h each, the FU2 tasks will be single "atomic tasks", which runtime is currently estimated at about 12h, so 25% shorter than FU1.
The "Gravitational Wave search S6Bucket Follow-up #2" is planned to be launched later this week, and is planned to run another four months.
BM
PS: I'd like to keep the original thread for more or less official announcements about future plans of the E@H team, so I moved the book discussion to the "Science book reviews & recommendations" thread.
Does "top candidates" mean you've actually seen signals, or is it where you expect them to be?
It means that these are the most promising candidates for being actual signals, pending further investigation that we intend to do with this follow-up "run".
Yes, these are only a few tasks tho and will last only a few days.
The additional tasks come from several sources:
a) we lost some results (very few indeed) during a server crash earlier this year and these are recomputed now.
b) we discovered that a batch of tasks had not been created at all in the first place because of a problem when running the workunit generator. Those are now computed for the first time.
c) The "UB" in S6BucketFU1UB stands for "undisturbed bands". The idea is that after visual inspection of plots generated from the raw results, we excluded certain results of the original S6Bucket run from the follow-up run because they appeared to be affected by instrumental artifacts or other disturbances and might need special treatment. Now we had another look at the results judged "potentially mildly disturbed" and found many that we now, after a closer look, think are actually not disturbed, so they can be followed up without any special post-processing.
All in all I think roughly 40k additional workunits were generated this way.
Thanks Ben, I have my missing
)
Thanks Ben, I have my missing bits now. A real signal may still line up in the bins after some de-dispersion attempt, now in 8-bit dynamic range and also remembering that the pulsar emissions are fairly broad across the bands of interest anyway. So that power, added in quadrature, if periodic in the time series can then be caught by Fourier etc. A good statistic plus details from that goes toward new observations to confirm and characterise maybe a new discovery. Actually that makes the process and its proven successes even more impressive! :-) :-)
One other question if I may : what's the order of magnitude of a 'typical' dispersion delay ( or the difference b/w that for lowest frequency of interest c/w the highest ) compared to the sampling interval ? Or if you like : what's the general range of how many samples do we shift the data across to successfully get the 'right' emission spectrum ? Hopefully I have worded that right .... ie. I don't know how to convert a delta_DM value to a time delay.
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
Hi Mike, RE: One
)
Hi Mike,
The dispersion delay between two radio frequencies f1 and f2 is
delta_t = 4149 secs * (f1^-2 - f2^-2) * DM,
where DM is measured in pc/cm^3 and f1 and f2 are measured in MHz. You can divide this by sampling time to get the dispersion time delay delta_t in samples. You can find this formula for example on this handy NRAO webpage. It's also equation (2.43) in the introduction of my PhD thesis.
For the PMPS data we have f1= 1230 MHz, f2= 1518 MHz, and for a DM of, say, 200 pc/cm^3 I get delta_t = 0.188 seconds, which corresponds to 753.5 of the 250-microsecond samples.
Cheers,
Benjamin
Einstein@Home Project
Thank you Ben. Hmmm ..
)
Thank you Ben. Hmmm .. typically hundreds of samples indeed.
Cheers, Mike.
( edit ) Something's broken with your PhD thesis link @ Harvard but I found this one.
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: Hi all, the following
)
Thank you much for the very helpful information regarding the new binary radio pulsar search.
Jeroen
As announced earlier the
)
As announced earlier the successor to the "S6Bucket Follow-up #1" (FU1) will be another follow-up on the candidates we got from the S6Bucket search, zooming even deeper into the top candidates from the "Follow-up #1" (FU2). Whereas the FU1 tasks consisted of 8 independent "atomic tasks" bundled together with an (initially rough) estimated runtime of 2h each, the FU2 tasks will be single "atomic tasks", which runtime is currently estimated at about 12h, so 25% shorter than FU1.
The "Gravitational Wave search S6Bucket Follow-up #2" is planned to be launched later this week, and is planned to run another four months.
BM
PS: I'd like to keep the original thread for more or less official announcements about future plans of the E@H team, so I moved the book discussion to the "Science book reviews & recommendations" thread.
BM
Does "top candidates" mean
)
Does "top candidates" mean you've actually seen signals, or is it where you expect them to be?
RE: Does "top candidates"
)
It means that these are the most promising candidates for being actual signals, pending further investigation that we intend to do with this follow-up "run".
BM
BM
There is new tasks available
)
There is new tasks available for S6 FU1UB?
RE: There is new tasks
)
Yes, these are only a few tasks tho and will last only a few days.
The additional tasks come from several sources:
a) we lost some results (very few indeed) during a server crash earlier this year and these are recomputed now.
b) we discovered that a batch of tasks had not been created at all in the first place because of a problem when running the workunit generator. Those are now computed for the first time.
c) The "UB" in S6BucketFU1UB stands for "undisturbed bands". The idea is that after visual inspection of plots generated from the raw results, we excluded certain results of the original S6Bucket run from the follow-up run because they appeared to be affected by instrumental artifacts or other disturbances and might need special treatment. Now we had another look at the results judged "potentially mildly disturbed" and found many that we now, after a closer look, think are actually not disturbed, so they can be followed up without any special post-processing.
All in all I think roughly 40k additional workunits were generated this way.
Cheers
HB
RE: All in all I think
)
Then you have to fix the status page for FU1. Currently it looks like we returned more results than we have received.