Comments (6)
Yes, this is expected behavior. And the quick reason why is that we are missing some data to be able to output the full-duration of the data because you are de-dispersing.
Since you are dedispering to a positive DM, you are assuming that the low-frequency data is arriving later than the high-frequency data. That's fine at the beginning of the observation where we can simply grab later samples at lower freqs to add them to the initial sample at the highest freq. But that doesn't work at the end of the observation where we have no low freq data recorded! So since we don't know what to add in order to properly dedisperse, we simply ignore parts of the original file where we don't have all frequencies we need.
This problem gets worse and worse (meaning you lose a bigger and bigger fraction of your input data) when 1) the observation duration is relatively short and 2) when the total amount dispersion (in time) is long between the top and bottom of the observing band.
In your case, for a DM of 300, the differential delay across the band (1200-1800MHz) is 0.48 seconds, and so you will always lose at least that amount of data.
However, there is another issue in that prepsubband (and prepdata) work on blocks of data of (by default) 2400 samples (see the Spectra/Subint number in the output) for filterbank data. So you will often get an integer number of samples of that duration out.
And finally, both of those commands automatically choose a highly factorable number of data points to output so that FFTs on that data will be efficient.
For your initial case you are seeing all of these effects. You can mitigate some of this (but you can't get back data you don't have because of dispersion) by specifying the number of output data points to use with the -numout
option. If you specify -numout
to equal the number of input data points, trying looking at the end of the resulting .dat file with the exploredat
command and you will see how the noise in the data changes as we have less and less data available to do the full and proper dedispersion.
Hope this helps!
from presto.
Thank you very much for your response.
I understood your explanation and it makes sense. I tried your suggestion of looking into the .dat file but I don't see anything weird. Is there a way to predict how much data I will lose? Because like you said, with the DM and frequency range I can obtain the least amount of data that I will lose, in the example you mentioned it was ~0.5 seconds but I am losing around a minute of data. I assume this is due to the integer number of blocks that you mentioned. When executing prepsubband
it says that for my original number of samples the "good number of samples" to work with is 30240 but the data written only gets to 24000. Is there a way to know how many blocks I will lose?
from presto.
Have you tried specifying -numout
? And where does it say "30240"?
from presto.
Yes I have tried specifying -numout
to be equal to the original samples.
I changed the PRESTO version to a more recent one so now the outputs look like this:
-prepsubband -nobary -lodm 300 -dmstep 1 -numdms 1 -downsamp 1 -nsub 2048 -runavg -numout 29868 -o prep_output 2022-09-10_11_15_25.404305.fil
And this is without the -numout
(where the 30240 comes from):
As you can see, the data points written in both cases is 24000 and is padding the rest, here is a screenshot of exploredat
of whren the -numout
is specified but it looks the same when is not specified.
from presto.
One other request for you to try: can you try using prepdata
to make a single time series rather than prepsubband
? The latter is quite a bit more complicated (since it allows you to output mutiple time series at once), so there might be a bit of extra overhead there.
from presto.
I tried prepdata
and it doesn't remove as much samples as prepsubband
. Specifying -numout
or not, it writes 28800 samples instead of the 24000 from prepsubband
.
This amount of data lost makes more sense giving the amount of samples per data block and the least amount of data I should lose in the process. So it seems that prepsubband
may be removing more samples in the process to optimize the output of many time series? In that case prepdata
seems to be the best command for my case.
Thank you very much for your help :)
from presto.
Related Issues (20)
- Packaging `PRESTO` for multiple OSes? HOT 3
- ld: cannot find -lfftw3f: No such file or directory HOT 2
- undefined symbol: _gfortran_st_close HOT 2
- PFD file format? HOT 4
- Problem installing PRESTO with Apple M2 - Ventura 13.2 HOT 2
- Remove the Pusar Miner in FAQ HOT 5
- a memory issue about test_presto_python HOT 4
- ERROR: Failed building wheel for pyslalib HOT 3
- ERROR:makewisdom.c:4:10: fatal error: meminfo.h: No such file or directory HOT 1
- Discrepancy between time for single pulse with single_pulse_search.py and watefaller.py HOT 1
- Installing `PRESTO`'s Python bindings fails for Python 3.12. HOT 4
- Doc fix HOT 2
- Feature request related to plotting HOT 10
- how to INSTALL TEMPO2 HOT 1
- Installation Issue: Wisdom not being made correctly (for me) HOT 2
- Bug? In `DDplan.py`, not actually using user imputed `--loDM` & `--hiDM` Or maybe I am not using the arg correctly? HOT 3
- Determining prepfold P and Pdot step sizes HOT 6
- prepfold runs many processes regardless of -ncpus setting HOT 4
- Question on the arrival time listed for a pulse candidate found in `single_pulse_search.py` HOT 1
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from presto.