psgeom

psgeom aims to provide an easy to use code base for common geometrical operations during scattering experiments:

• load/save multiple geometry formats
• translate/rotate parts of a detector
• easily compute reciprocal/polar space coordinates
• perform angular integration

User-friendliness is emphasized. The software aims to be general but not sacrifice simplicity.

Chances are that if you need to compute stuff relating to scattering geometry for a particular experiment, psgeom has what you need, and it will be easy to use.

Currently, interfaces exist to geometry formats from:

• psana
• CrystFEL
• cheetah
• DIALS
• LCLS detector group metrologies
• a simple flat text / HDF5 pixel map

TJ Lane [email protected]

---

Scripts

A lot of users just want to convert geometry files between different formats, or other simple tasks. To assist, psgeom provides a few command line scripts that will be of general interest:

• geoconv : convert between geometry file formats
• geoQ : quickly get reciprocal space coordinates for a geometry
• gainmk and gainconv : for CSPAD gain files -- to make a new one and convert it's formatting, respectively

Run any script with a -h flag to get more information

---

Examples

A few examples of how to use the code.

format conversions

from psgeom import camera

geom = camera.CompoundAreaCamera.from_psana_file('1-end.data')
geom.to_crystfel_file('my_new.geom')

looking at pixel positions

geom = camera.CompoundAreaCamera.from_crystfel_file('my.geom')
print(geom.xyz) # real-space xyz coords

looking at things in basisgrid format

By "basisgrid", we mean a the geometry is described as a set of panels; each panel by a vector pointing to the first pixel to be read from memory, along with two vectors for the slow/fast scan directions.

bg = geom.to_basisgrid()
for g in range(bg.num_grids):
    print(bg.get_grid(g))

radial averaging

from psgeom import bin

xyz = geom.xyz

beam_vector = np.array([0.0, 0.0, 1.0])     # assumed
wavenumber = 2.0 * np.pi / wavelength       # inv A

norm = np.linalg.norm(xyz, axis=-1)
S = xyz / norm[...,None] # unit vector

q_xyz = wavenumber * (S - beam_vector)
q_mag = np.linalg.norm(q_xyz, axis=-1)

radavg = bin.Averager(q_mag, mask, n_bins=500)

# data is raw detector data, Iq is radial average in q-coords
Iq1 = radavg(data1) 
Iq2 = radavg(data2) 
...

an easier way to compute reciprocal coords

from psgeom import camera
from psgeom import reciprocal

geom = camera.CompoundAreaCamera.from_crystfel_file('my.geom')
eV = 9500.0
d = reciprocal.Geometry(geom, eV)

d.xyz        # real space cart
d.polar      # real space polar
d.reciprocal # reciprocal cart
d.recpolar   # reciprocal polar coords

# compute interpolated values at specific intersection points
pix, intersect = d.compute_intersections(interp_vectors, 0) # 0 --> grid_index

---

How it works

In the scattering community, there are two principle paradigms in use for representing complex detector/experimental geometries: the "basisgrid" paradigm, and the "elemental" paradigm. psgeom implements both, which means it is both extremely powerful and capable of converting between formats that use either paradigm.

The "basisgrid" paradigm represents a geometry as a set of 2d planar sensors in 3d space. These sensors are represented by a position (p) vector that points to the first pixel of that panel to be read from memory. Two additional vectors correspond to the slow (s) and fast (f) scan directions, showing how to map a 2d array of data onto the sensor geometry. Used by, for example, crystFEL.

The "elemental" paradigm defines sensor components by hand, and then builds a more complicated sensor geometry by translating/rotating these elements with respect to one another, perhaps in a heirarchical fashion. This provides a bit extra power in terms of representing how rigid units move in space, as they are often correlated (e.g. mounted on a hard platform driven by motors in a beamline). Because the elements must be specified in software, this paradigm comes with additional complexity.

In reality all known x-ray and electron detectors to date (as of April 2020) are , composed of planar, 2d arrays of rectangular pixels. psgeom reflects this by emphasizing element definitions that correspond to this pattern; such elements are defined by their (1) pixel shape, (2) array size (n-by-m pixels), and (3) any gaps in the sensor surface. This allows for conversion between the "elemental" and "basisgrid" paradigms.

A paper describing these formats and the mathematics behind them is in preparation

---

Quick Links

Information about the psana geometry: https://confluence.slac.stanford.edu/display/PSDM/Detector+Geometry

List of high quality geometries generated by users and optical metrologies generated by LCLS: https://confluence.slac.stanford.edu/display/PSDM/Geometry+History

CrystFEL Geometry: http://www.desy.de/~twhite/crystfel/manual-crystfel_geometry.html