It is not unlikelly that I go to bed at those hours, but I live in Chile, and here right now is just 10pm
You don't have to worry about the pedestal. Latest versions of DBE included a section, at the bottom, about model application. To simplify, enable the Normalize option (this will calculate and apply the pedestal automatically), discard the model, and apply DBE to the target image. Now, to compare results, just have to undo/redo. Not as fast as with a preview, but helps a lot. You may create duplicates of the image with the best results so far, to compare with other set of parameters/samples.
(I've been here for a while, so sometimes I forget the latest additions
)
A quick description of what is that famous pedestal: Think a bit on the flat frame calibration scheme. You want to correct the data information from uneven illumination from optical design, dust motes, and also different pixel response. So, you create a "model", by imaging a uniform screen (or by any other method). The result is a image, with pixel values. If you perform a straight division, you modify the flux of the entire image. As a trivial experiment, think that every pixel is divided by 0.25 (no reason for that particular number). It is equivalent to multiplying the image by 4... everything changes. So, if you want to keep the image as closely as you can to the original state, in terms of flux, you must normalize the flat frame. Usually this is done by dividing it by a constant number, like the median of its pixels, or the average. Or, what is the same, dividing the light frame by the "raw" flat, and then multiplying by it's median/average/other value. The same can be said for additive corrections (sky subtraction, from light pollution gradients or other effects). To keep the original flux constant, we have to normalize the model, or add a pedestal, witch is calculated in the same way.
WARNING: the real background has noise, small scale variations and other elements that make it not as smooth as the model. A straight subtraction, without pedestals or rescalations will yield in a heavy data loss. That operation is in some way equivalent to a very aggressive black point setting, at the Histogram, clipping all the data below the maximum of the "data bell".
Back on topic, current DBE's and ABE's implementations do the work for you, so forget all that PM stuff
PS: Nice image! That reminds me to upload mine
Tomorrow... maybe
