Software crash when creating master bias

[-Amenophis-]

Good evening everyone,

I currently have a problem with PI regarding the generation of a master BIAS.

I have about 600 bias so I use the default settings are:

 "buffer size: 32"
  "Stack size: 2048"

I use an iMAC with 8GB of RAM DDR3 1600MHz.

Launching the process icon with the algorithm "Winsorized Sigma Clipping" all goes well then PI plant:

[IMG] http://i.imgur.com/A0mzabX.png [/ IMG]

Then I reduce both values??:

 "buffer size: 16"
  "Stack size: 1024"

IDEM

Then with:

 "buffer size: 8"
  "Stack size: 512"

IDEM also.

So I tried with 100 bias and there, it works nickel: b:

I try with  200 bias  ...... and it crashes. O

With the algorithm "Linear Fit clipping" with the basic parameters, PI also plant.

What's the worry?

Thank you in advance for your help zamis;)

[-Amenophis-]

Someone he try this technique ?

Try a hierarchical approach: to integrate 1000 images, you can integrate your images in groups of 200, and then, in a second step, stack those 5 images. The result should be the same -except for pixel rejection. But groups of 200 images should already have stable enough statistics for good pixel rejection.

[pfile]

as you might have read, this is a 'bug' in OSX. the process limit for open files is around 200 or so. not much that can be done other than to try the hierarchical approach or boot windows/linux.

rob

[Juan Conejero]

As Rob has said, Mac OS X limits the total number of open files to 256. This means that you cannot integrate more than about 200 images in a single run of ImageIntegration. We haven't found a valid (stable) solution to this problem yet on OS X. For integration of large data sets with PixInsight, we recommend using FreeBSD, Linux or Windows. See for example this thread for a report similar to yours.

Try a hierarchical approach: to integrate 1000 images, you can integrate your images in groups of 200, and then, in a second step, stack those 5 images. The result should be the same -except for pixel rejection.

This can be a workaround, but pixel rejection will always work better with a larger set, especially if you use one of the more sophisticated algorithms such as linear fit rejection or Winsorized sigma clipping.