CrystFEL detector geometry description files

The detector geometry is taken from a text file rather than hardcoded into the program. Programs which care about the geometry (particularly indexamajig and pattern_sim take an argument --geometry=filename, where filename contains the geometry. See the "examples" folder for some examples (look at the ones ending in .geom).

A flexible (and pedantic) representation of the detector has been developed to avoid all possible sources of ambiguity. CrystFEL's representation of a detector is broken down into one or more "panels", each of which has its own camera length, geometry, resolution and so on. Each panel fits into an overall data block taken from the HDF5 file, defined by minimum and maximum coordinates in the "fast scan" and "slow scan" directions. "Fast scan" refers to the direction whose coordinate changes most quickly as the bytes in the HDF5 file are moved through. The coordinates are specified inclusively, meaning that a minimum of 0 and a maximum of 9 results in a width of ten pixels. Counting begins from zero. All pixels in the image must be assigned to a panel - gaps are not permitted.

The job of the geometry file is to establish a relationship between an array of pixel values in the HDF5 file, defined in terms only of the "fast scan" and "slow scan" directions, and the laboratory coordinate system defined as follows:

Naively speaking, this means that CrystFEL looks at the images from the "into the beam" perspective, but please avoid thinking of things in this way. It's much better to consider the precise way in which the coordinates are mapped.

The geometry file should contain lines of the following form:

panel/clen = 0.560

panel can be any name of your choosing. You can make up names for your panels however you please, as long as the first three letters are not "bad" or "rigid_group" (in lower case), because these are used for special purposes (see below).

You can also specify values without a panel name, for example:

clen = 0.560

In this case, the value will be used for all subsequent panels appearing in the file which do not have their own specific values for the property, or until you specify another default value further down the file. Panel-specific values always have priority over default values, and changing the default value has no effect for panels which had already be mentioned at the point in the file where the default value was specified.

Comments start with a semicolon. They will be ignored.

Top-level only parameters

The parameters in this section can only appear without a panel name.

wavelength nnn [m|A]
photon_energy nnn [eV|keV]
electron_voltage nnn [V|kV]

These statements specify the incident radiation wavelength. You must include one (not more) of these statements. wavelength specifies the wavelength directly, photon_energy specifies the energy per photon for electromagnetic radiation (e.g. X-rays), and electron_voltage specifies the accelerating voltage for an electron beam.

nnn can be a literal number, or it can be a header location in the image data file. In the latter case, the program will do what you expect in the case of multi-frame data files, e.g. a scalar value in the metadata will be applied to all frames, or an array of values can be used to provide a separate wavelength for each frame.

Units should be specified after the value (or location). These can be m or A for wavelength, eV or keV for photon_energy, and V or kV for electron_voltage. For photon_energy, if no units are given then the value will be taken to be in eV.

detector_shift_x = nnn [m|mm]
detector_shift_y = nnn [m|mm]
These specify that the entire detector should be shifted by this amount in the x and y directions. The units should be specified as m or mm. If units are not specified, the value will be taken as metres. nnn can be a file metadata location (e.g. an HDF5 path).
peak_list = loc
This gives the location of the peak list in the data files, for peak detection methods hdf5 and cxi (see man indexamajig).
peak_list_type = layout

Specify the layout of the peak list. Allowed values are cxi, list3 and auto.

list3 expects the peak list to be a two dimensional array whose size in the first dimension equals the number of peaks and whose size in the second dimension is exactly three. The first two columns contain the fast scan and slow scan coordinates, the third contains the intensities. This is the correct option for "single-frame" HDF5 files as written by older versions of Cheetah.

cxi expects the peak list to be a group containing four separate HDF5 datasets: nPeaks, peakXPosRaw, peakYPosRaw and peakTotalIntensity. See the specification for the CXI file format at for more details. This is the correct option for "multi-event" HDF5 files as output by recent versions of Cheetah.

auto tells CrystFEL to decide between the above options based on the file extension.

Note that CrystFEL considers all peak locations to be distances from the corner of the detector panel, in pixel units, consistent with its description of detector geometry (see 'man crystfel_geometry'). The software which generates the HDF5 or CXI files, including Cheetah, may instead consider the peak locations to be pixel indices in the data array. To compensate for this discrepancy, CrystFEL will, by default, add 0.5 to all peak coordinates. Use --no-half-pixel-shift if this isn't what you want.

Per-panel values

The following parameters can be set for each panel individually. Don't forget that they can also be used at the "top level" to set default values.


The location in the HDF5 file of the data block that contains the panel's data. The default value is /data/data. If the HDF5 file contains multiple events, and each event is stored in a different data block, the variable part of the path can be represented using the % character placeholder. Example:

data = /data/%/rawdata

The CrystFEL programs will look for the first event at /data/event1_name/rawdata, for the second at /data/event2_name/rawdata, etc., where event_name event2_name are simply whatever the program could find in the HDF5 file which matched the pattern you gave.


Information about the layout of the data block identified by the 'data' property. n is an integer number identifying an axis in a multidimensional HDF5 data block. The property value defines the kind of information encoded by the axis. Possible values are:

event placeholder,the axis encodes events
the axis encoding the slow scan index
the axis encodes the fast scan index
the index in this dimension should be fixed at 'number'.

CrystFEL assumes that the data block defined by the 'data' property has a dimensionality corresponding to the axis with the highest value of n defined by the 'dim' property. That is, if the geometry file specifies dim0, dim1 and dim2, then the data block is expected to be three-dimensional. The size of the data block along each of those axes comes from the image metadata (e.g. the array sizes in the HDF5 file).

The lowest number of n corresponds to the most slowly-changing array index as the data block is traversed. The default values are dim0=ss and dim1=fs. The value of n corresponding to fs must not be lower than the value assigned to ss, i.e. "fast scan is always fast scan".


dim0 = %
dim1 = 4
dim2 = ss
dim3 = fs

The above snippet specifies that the data block is 4-dimensional. The first axis represents the event number, the index in the second axis is always 4, and the remaining two axes are the image coordinates.

The range of pixels in the data block which correspond to the panel, in fast scan/slow scan coordinates, specified inclusively.
The number of detector intensity units (ADU) which will arise from either one electron-Volt of photon energy, or one photon. This is used to estimate Poisson errors. Note that setting different values for this parameter for different panels does not result in the intensities being scaled accordingly when integrating data, but it does affect the intensities calculated by pattern_sim. You should only specify one out of adu_per_eV and adu_per_photon.
The resolution (in pixels per metre) for this panel. This is one over the pixel size in metres.
The camera length (in metres) for this panel. You can also specify the HDF5 path to a scalar floating point value containing the camera length in millimetres. For example: "panel0/clen = /LCLS/detectorPosition". If the HDF5 file contains more than one event, and the data block is scalar, the camera length value it contains will be used for all events. If, however, the data block is multidimensional and the second dimension is bigger than one, the CrystFEL programs will try to match the content of the data block with the events in the file, assigning the first value in the data block to the first event in the file, the second value in the data block to the second event in the file, etc. See coffset as well.
The camera length offset (in metres) for this panel. This number will be added to the camera length (clen). This can be useful if the camera length is taken from the HDF5 file and you need to make an adjustment, such as that from a calibration experiment.
For this panel, the fast and slow scan directions correspond to the given directions in the lab coordinate system described above, measured in pixels. Example: "panel0/fs = 0.5x+0.5y-0.0001z". Including a component in the z direction means that the panel is not perpendicular to the X-ray beam.
The corner of this panel, defined as the first point in the panel to appear in the HDF5 file, is now given a position in the lab coordinate system. The units are pixel widths of the current panel. Note that "first point in the panel" is a conceptual simplification. We refer to that corner, and to the very corner of the pixel - not, for example, to the centre of the first pixel to appear.
The saturation value for the panel. You can use this to exclude saturated peaks from the peak search or to avoid integrating saturated reflections. However, usually it's best to include saturated peaks, and exclude saturated reflections with the --max-adu option of process_hkl and partialator. Therefore you should avoid setting this parameter - a warning will be displayed if you do.
Set this to 1 or "true" to ignore this panel completely.
Mark pixels as "bad" if their values are respectively less than, more than or equal to the given value. Note carefully that the inequalities are strict, not inclusive: "less than", not "less than or equal to".
Mark the specified number of pixels, at the edge of the panel, as "bad".

These specify the parameters for bad pixel mask number N. You can have up to 8 bad pixel masks, numbered from 0 to 7 inclusive. Placeholders ('%') in the location (maskN_data) will be substituted with the same values as used for the placeholders in the image data, although there may be fewer of them for the masks than for the image data.

+You can optionally give a filename for each mask with maskN_file. The filename may be specified as an absolute filename, or relative to the working directory. If you don't specify a filename, the mask will be read from the same file as the image data.

A pixel will be considered bad unless all of the bits which are set in goodbits are set. A pixel will also be considered bad if any of the bits which are set in badbits are set. Note that pixels can additionally be marked as bad via other mechanisms as well (e.g. no_index or bad).


mask2_data = /data/bad_pixel_map
mask2_file = /home/myself/mybadpixels.h5
mask2_goodbits = 0x00
mask2_badbits = 0xff

There are some older mask directives which are still understood by this version of CrystFEL. They are synonyms of the new directives as follows:

mask       ----->   mask0_data
mask_file  ----->   mask0_file
mask_good  ----->   mask0_goodbits
mask_bad   ----->   mask0_badbits
This specifies the location of the per-pixel saturation map in the HDF5 file. This works just like mask in that it can come from the current file or a separate one (see saturation_map_file). Reflections will be rejected if they contain any pixel above the per-pixel values, in addition to the other checks (see max_adu).
Specifies that the saturation map should come from the HDF5 file named here, instead of the HDF5 file being processed. It can be an absolute filename or relative to the working directory.

Bad regions

Bad regions will be completely ignored by CrystFEL. You can specify the pixels to exclude in pixel units, either in the lab coordinate system (see above) or in fast scan/slow scan coordinates (mixtures are not allowed). In the latter case, the range of pixels is specified inclusively. Bad regions are distinguished from normal panels by the fact that they begin with the three letters "bad".

If you specify a bad region in fs/ss (image data) coordinates, you must also specify which panel name you are referring to.

Note that bad regions specified in x/y (lab frame) coordinates take longer to process (when loading images) than regions specified in fs/ss (image data) coordinates. You should use fs/ss coordinates unless the convenience of x/y coordinates outweighs the speed reduction.


badregionA/min_x = -20.0
badregionA/max_x = +20.0
badregionA/min_y = -100.0
badregionA/max_y = +100.0
badregionB/min_fs = 128
badregionB/max_fs = 160
badregionB/min_ss = 256
badregionB/max_ss = 512
badregionB/panel = q0a1

See the "examples" folder for some examples (look at the ones ending in .geom).

Rigid groups and rigid group collections

Some operations in CrystFEL, such as refining the detector geometry, need a group of panels to be treated as a single rigid body. Such "rigid groups" might describe the fact that certain panels are physically connected to one another, for example, a pair of adjacent ASICs in the CSPAD detector. Rigid groups can be defined in the geometry file by listing the panels belonging to each group and assigning the group a name, like this:

rigid_group_myrg = panel1,panel2

This creates a rigid group called myrg, containing panels panel1 and panel2.

You can specify multiple sets of rigid groups. For example, as well as specifying the relationships between pairs of ASICs mentioned above, you may also want to specify that certain groups of panels belong to an independently-movable quadrant of the detector. You can declare and name such "rigid group collections" as follows:

rigid_group_collection_myrgcoll = rigidgroup1,rigidgroup2

This creates a rigid group collection called myrgcoll, containing rigid groups rigidgroup1 and rigidgroup2.

Definitions of rigid groups and rigid group collections can appear at any place in the geometry file and can be declared using the following global properties. They are not panel properties, and therefore don't follow the usual panel/property syntax. You can assign any number of panels to a rigid group, and any number of rigid groups to a rigid group collection. A panel can be a member of any number of rigid groups.


This page was written by Thomas White.

Reporting bugs

Report bugs to

Copyright and Disclaimer

Copyright © 2012-2021 Deutsches Elektronen-Synchrotron DESY, a research centre of the Helmholtz Association.

Please read the AUTHORS file in the CrystFEL source code distribution for a full list of contributions and contributors.

CrystFEL is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

CrystFEL is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

You should have received a copy of the GNU General Public License along with CrystFEL. If not, see

See also

crystfel pattern_sim indexamajig

If CrystFEL is installed on your computer, you can read this manual page offline using the command man crystfel_geometry.