#!/usr/bin/env python # -*- coding: utf-8 -*- ''' Created on 10 Apr 2013 @author: Éric Piel This is a simple example on how to acquire a "spectrum cube" on the SPARC in a script. Run as: cl_acquisition output.h5 You first need to run the odemis backend with the SPARC config: odemisd --log-level 2 install/linux/usr/share/odemis/sparc-amolf.odm.yaml To change some configuration settings, you can use the cli: # to change the yaw of the mirror (by 10 um): odemis-cli --move MirrorMover rz -10 # to change the center bandwidth of the spectrograph (to 520 nm): odemis-cli --position SP2300i wavelength 0.520 # to change the resolution of the SEM (i.e., number of points acquired): odemis-cli --set-attr "EBeam ExtXY" resolution "20, 10" # to change the exposure time of the spectrometer (to 100 ms): odemis-cli --set-attr Spec10 exposureTime 0.1 # to change the binning of the Angle-Resolved camera: odemis-cli --set-attr ARCam binning "2, 2" ''' from __future__ import division import logging import numpy from odemis import model, dataio from odemis.model import MD_PIXEL_SIZE, MD_POS import sys import threading import time logging.getLogger().setLevel(logging.DEBUG) # region of interest as left, top, right, bottom (in ratio from the whole area) RECT_ROI = (0.25, 0.25, 0.75, 0.75) class Acquirer(object): def __init__(self): # the ebeam scanner self.escan = None # the secondary electron detector self.sed = None # the spectrometer self.spect = None # find components by their role for c in model.getComponents(): if c.role == "e-beam": self.escan = c elif c.role == "se-detector": self.sed = c elif c.role == "spectrometer": self.spect = c if not all((self.escan, self.sed, self.spect)): logging.error("Failed to find all the components") raise KeyError("Not all components found") self.set_roi(RECT_ROI) self.acq_left = 0 # how many points left to scan self.acq_spect_buf = [] # list that will receive the acquisition data (numpy arrays of shape (N,1)) self.acq_complete = threading.Event() def set_roi(self, rect): # save the resolution res = self.escan.resolution.value center = ((rect[0] + rect[2]) / 2, (rect[1] + rect[3]) / 2) width = (rect[2] - rect[0], rect[3] - rect[1]) shape = self.escan.shape scale = (1 / width[0], 1 / width[1]) # translation is distance from center (situated at 0.5, 0.5), can be floats trans = (shape[0] * (center[0] - 0.5), shape[1] * (center[1] - 0.5)) # always in this order self.escan.scale.value = scale # might update the resolution & translation self.escan.resolution.value = res self.escan.translation.value = trans def acquire_cube(self): """ acquires two images: the SEM secondary electron image and the spectrum cube return (tuple of numpy.array): sed_data spect_data """ exp = self.spect.exposureTime.value #s sem_size = self.escan.resolution.value spect_size = self.spect.resolution.value assert spect_size[1] == 1 # it's supposed to be a band numberp = numpy.prod(sem_size) # magical formula to get a long enough dwell time. # works with PVCam, but is probably different with other drivers :-( readout = numpy.prod(spect_size) / self.spect.readoutRate.value + 0.01 self.escan.dwellTime.value = (exp + readout) * 1.1 + 0.05 # 50ms to account for the overhead and extra image acquisition # pixel write/read setup is pretty expensive ~10ms expected_duration = numberp * (self.escan.dwellTime.value + 0.01) logging.info("Starting acquisition of about %g s...", expected_duration) self.acq_spect_buf = [] self.acq_left = numberp self.acq_complete.clear() # synchronize the two devices self.spect.data.synchronizedOn(self.escan.newPosition) startt = time.time() self.spect.data.subscribe(self.receive_spect_point) sed_data = self.sed.data.get() # wait the last point is fully acquired self.acq_complete.wait() endt = time.time() logging.info("Took %g s", endt - startt) self.spect.data.synchronizedOn(None) # create metadata for the spectrum cube from the SEM CL # Could be computed also like this: # eps = self._emitter.pixelSize.value # scale = self._emitter.scale.value # ps = (eps[0] * scale[0], eps[1] * scale[1]) # md = {MD_PIXEL_SIZE: ps} md_sem = sed_data.metadata md = {} for m in [MD_PIXEL_SIZE, MD_POS]: if m in md_sem: md[m] = md_sem[m] # create a cube out of the spectral data acquired # dimensions must be wavelength, Y, X assert len(self.acq_spect_buf) == numberp # each element of acq_spect_buf has a shape of (1, N) # reshape to (N, 1) for e in self.acq_spect_buf: e.shape = e.shape[-1::-1] # concatenate into one big array of (N, numberp) spect_data = numpy.concatenate(self.acq_spect_buf, axis=1) # reshape to (N, Y, X) spect_data.shape = (spect_size[0], sem_size[1], sem_size[0]) # copy the metadata from the first point md_spect = self.acq_spect_buf[0].metadata md_spect.update(md) spect_data = model.DataArray(spect_data, metadata=md_spect) return sed_data, spect_data def receive_spect_point(self, dataflow, data): """ callback for each point scanned as seen by the spectrometer """ self.acq_spect_buf.append(data) self.acq_left -= 1 if self.acq_left <= 0: dataflow.unsubscribe(self.receive_spect_point) self.acq_complete.set() if __name__ == '__main__': # must have one argument as name of the file which contains SEM and spectrogram acquisitions if len(sys.argv) != 2: logging.error("Must be called with exactly 1 argument") exit(1) filename = sys.argv[1] acquirer = Acquirer() sed_data, spect_data = acquirer.acquire_cube() # to tell the exporter that the 3rd dimension of the spectrum is the channel # it has to be the 5th dimension => insert two axes s = spect_data.shape spect_data.shape = (s[0], 1, 1, s[1], s[2]) exporter = dataio.find_fittest_converter(filename) exporter.export(filename, [sed_data, spect_data])