Treatment of non-uniformly sized pixels

[graeme-winter]

For Eiger detectors there are double-sized and quad-sized pixels, which are decomposed in the hardware into multiple pixels with similar pixel values (i.e. 7 -> 3+4 with some random element used to assign the spare count). This is (i) expensive and (ii) makes the data an annoying size e.g. 512x1028 pixels from 8 x (256x256) pixel ASICs.

Treatment of "big" pixels in spot finding would be simple, as we could just add 2, 4 to the N mask SAT in place of 1, but wider handling is likely to be ... annoying.

Questions:

• can we handle this gracefully w.r.t. the mm to pixel, pixel to mm conversions?
• can we add support for non-uniform pixel sizes to spot finding as sketched above?
• how would / could we handle this in integration

Motivation:

• Jungfrau data requires similar treatment which is expensive in the acquisition chain
• performance could potentially be improved if we treat the data in the same shape it was measured in
• it is the truth

[graeme-winter]

For this I think it is reasonable to consider only pixels with size 1, 2, 4 x some fundamental size.

[dagewa]

For this I think it is reasonable to consider only pixels with size 1, 2, 4 x some fundamental size.

To register a vote for more general handling, Timepix-based detectors often have a central cross that is composed of pixels that are three times wider than a normal pixel.

[graeme-winter]

For this I think it is reasonable to consider only pixels with size 1, 2, 4 x some fundamental size.

To register a vote for more general handling, Timepix-based detectors often have a central cross that is composed of pixels that are three times wider than a normal pixel.

Noted. May we constrain this to natural numbers?

[biochem-fan]

Do output files contain raw (original) values of these big pixels? I thought files (at least from PILATUS and EIGER) contain values re-distributed to a uniform grid of (virtual) pixels. Are you going to sum values to get the original value on a big physical pixel? But the re-distribution algorithm is unknown to us and resummation does not necessarily give us the original value, does it?

[graeme-winter]

Do output files contain raw (original) values of these big pixels? I thought files (at least from PILATUS and EIGER) contain values re-distributed to a uniform grid of (virtual) pixels. Are you going to sum values to get the original value on a big physical pixel? But the re-distribution algorithm is unknown to us and resummation does not necessarily give us the original value, does it?

It should give the original value for Eiger 2, Pilatus 3 - but my motivation is more for treatent of the uncorrected data or for data from jungfrau

[graeme-winter]

@biochem-fan I can see if I can get a data set collected with virtual_pixel_correction_applied=false [1] to use as a reference case for this

[1] https://media.dectris.com/SIMPLON_APIReference_v1p8.pdf