HDFCatalog.to_mesh generate negative values for the density field 'delta+1'.

[WangYun1995]

By using HDFCatalog.to_mesh, I want to generate a mesh of 1024^3 grids from the discrete particles of TNG100 simulation.

in_mesh = HDFCatalog.to_mesh(Nmesh=1024,BoxSize=75,dtype='f8', interlaced=True,
                             compensated=True, resampler='cic', position='Coordinates',weight='Masses')

However, I find that there are many negative values, which is wrong because that delta+1>=0.
How to fix this issue?

[sjforeman]

I can't remember encountering this issue when I was working with TNG. Could you post more of the code you're using to read in the TNG snapshot and make the mesh? Also, does the same issue happen when you use the following script to make a mesh? https://github.com/sjforeman/bskit/blob/master/scripts/grids/convert_illustris_hdf5_particles_to_bigfile_grid.py

[WangYun1995]

When I use this script <https://github.com/sjforeman/bskit/blob/master/scripts/grids/convert_illustris_hdf5_particles_to_bigfile_grid.py>, i.e. set 'compensated=Flase', everything is Okay: Min(delta+1)=0. Simon Foreman ***@***.***> 于2023年2月21日周二 02:58写道：

…

I can't remember encountering this issue when I was working with TNG. Could you post more of the code you're using to read in the TNG snapshot and make the mesh? Also, does the same issue happen when you use the following script to make a mesh? https://github.com/sjforeman/bskit/blob/master/scripts/grids/convert_illustris_hdf5_particles_to_bigfile_grid.py — Reply to this email directly, view it on GitHub <#676 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AQIHH56XIYXPMRO2AHQBA4DWYO5FTANCNFSM6AAAAAAVANP6KE> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[sjforeman]

I see. Does the issue appear if you use the script, but change it to use compensated=True?

[WangYun1995]

Yes, the same issue still exists. Simon Foreman ***@***.***> 于2023年2月21日周二 08:17写道：

…

I see. Does the issue appear if you use the script, but change it to use compensated=True? — Reply to this email directly, view it on GitHub <#676 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AQIHH532GLT4TIZJNTG5TGLWYQCSDANCNFSM6AAAAAAVANP6KE> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[rainwoodman]

It is not as surprising that the density can get negative after compensation. The compensation (mostly in the fourier space) produces a non-physical field that is optimized for power spectrum estimation.

If your next modelling step requires a positive definite density, then disable compensation(and interleaving) is the way to go. The effect mostly affects the small scale behavior of the field.

[WangYun1995]

So when I perform the wavelet transform of the density field, I need to disable compensation.