-
Notifications
You must be signed in to change notification settings - Fork 16
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Imaging with Powderday #137
Comments
Hi, I have switched to 661c31d
|
hi - can you possibly make a copy of the snapshot you're trying to run available somewhere, along with parameter files? |
oh I know the issue here - it's that the issues fixed by commit 224e78a are cropping up now. I'll test the imaging myself to see if I can understand If there's a bug. |
hi - I think I may have fixed this...can you please pull from master (952af33) and see if this works? |
Hi - I have obtained a new file called "convolved.134.hdf5" (related to gizmo_mw) as expecting with your quickstart: https://powderday.readthedocs.io/en/latest/quickstart.html#imaging Just need to see if it also works with my own snapshots. |
Hi - beginner questions:
but my output file is empty. Whereas it contains
|
Hi - I have a question about how pd performs simulation with filters.
It takes a really decent time to compute. However, if I ask for multiple filters like:
or
Both jobs are still not completed after 100+ hours of computation. Is that normal? |
Update:
|
Hi - very sorry for the delay. The run typically shouldn't take a whole week though it is true that imaging is the slowest part of the code, and the more filters you add, it will add a roughly linear amount to the computation per wavelength in the filter files. So this could become quite onerous. Are you able to get the code to finish with just a very simple filter file that has only one or two wavelengths just to debug this? |
Hi - Thank you again for your time. Should I try to put |
What is your overall goal? That might help enable me to point you in the right direction. |
My overall goal first is to produce SED of isolated galaxy using my own snapshots (and possibly compare them with SKIRT's one) and also to produce "realistic" RGB images of these galaxies. e.g. Figure 10 of the Powderday's paper: https://arxiv.org/pdf/2006.10757.pdf |
Okay great! For the images -- Greg Snyder made these himself. I'm linking the codes he used and a short description he gave to me via email.
|
Hi - In the example of Greg Snyder, he used a file named
However, my simulations gave me the following:
Am I missing an option in |
oh sorry about that - these are just npz files I packaged for greg. they're just the 2D image array at a given wavelength -- i.e. they look like this:
|
Thanks for the clarification. |
Should I use
Because, by showing files, they contains
|
sorry - what I mean is:
will have in it the 2D array with the image in it. you could use this to create an npz file to use with Greg's code (or alternatively change up his code to use the 2D array information from the imaging HDF5 file -- i.e. basically merge what's happening in the make_image_single_wavelength.py file with the lines:
another way of saying this is: for the example in the powderday paper, I ran 3 images at 1 micron (r), 0.5 microns (g), and 0.3 microns (b) -- these are the 1, 0.5 and 0.3 in the file names above. then I saved the 2D arrays of those images (using the |
Hi - Thank you for your kind help and your time, really appreciated. I have decided to merge both codes into one (only using external make_color_image of Greg for the make_interactive_nasa function) The main core of my function (to compute r, g and b) is:
Is the computation of r,g,b with |
Hi, I think this should be right -- assuming that you have run an imaging run with a filter set up at 0.3, 0.5 and 1 micron. What looks strange about the 2D arrays? |
great glad we're getting somewhere good! a few things come to mind: a. first I think there is some tuning necessary in greg's code to get the image to be aesthetically pleasing. here, it looks like there's some saturation in the middle of the disk> b. I've found that increasing the photon count can have a dramatic impact on images. what photon count are you using? perhaps trying increasing by a factor 10 to see if it helps? |
I used 1e6 photons and moved to 1e7. I have modified parameters in greg's code but still obtained an image like Could this come from the filters used? |
Hi - It seems that 1e8 photons take quit a while to perform (still not finished). |
hi yes - lower resolution simulations with few particles can only be so refined and will therefore limit the resolution of the image |
Hi - Sorry to bother you but, do you have any idea where the noise is coming from, and if it would be possible to remove it? |
Hi - there are generally two possibilities here: (1) low photon count. 1e8 will be expensive but will solve whether or not increasing photon count makes things better can you describe a bit more the parent simulation? |
Hi - |
Hi, Ah yeah - so this may be somewhat limited by NGAS. You could see if you could run a simulation with a factor of 8 higher resolution possibly to see if it looked better -- if you could compare just an early few snapshots (at least, once enough metals have formed that you have enough dust since presumably your dust content is set by the dust to metals ratio) with this run that might be telling. I'm guessing this is an idealized disk so hopefully not too difficult to run? |
Hi, |
great keep me posted! |
I would like to produce images following the "Imaging" section of the website:
https://powderday.readthedocs.io/en/latest/quickstart.html#imaging
in order to obtain RGB images (like the figure 10 of the Powderday's paper: https://arxiv.org/pdf/2006.10757.pdf)
At first, I have set the following parameters in parameters_master:
The goal would be to use after another filter like GALEX one, but first I am using the arbitraty one.
When I execute powderday, I have the following issue:
The text was updated successfully, but these errors were encountered: