diff --git a/cal_tools/cal_tools/agipdlib.py b/cal_tools/cal_tools/agipdlib.py
index 9ee105b9993396873817c928bfa5ef12ee154f7a..7509813830d60edd8e646fac6634ae04b09c02e8 100644
--- a/cal_tools/cal_tools/agipdlib.py
+++ b/cal_tools/cal_tools/agipdlib.py
@@ -251,6 +251,10 @@ class AgipdCorrections:
self.corr_bools.get('blc_stripes'),
self.corr_bools.get('melt_snow')]
+ self.blc_bools = [self.corr_bools.get('blc_noise'),
+ self.corr_bools.get('blc_hmatch'),
+ self.corr_bools.get('blc_stripes')]
+
def read_file(self, i_proc, file_name):
"""
Read file with raw data to shared memory
@@ -433,39 +437,37 @@ class AgipdCorrections:
n_img = self.shared_dict[i_proc]['nImg'][0]
data = self.shared_dict[i_proc]['data'][:n_img]
cellid = self.shared_dict[i_proc]['cellId'][:n_img]
- mask = self.mask[module_idx] # shape of n_cells, x, y
+ mask_std = self.mask[module_idx] # shape of n_cells, x, y
for c in cells:
std = np.nanstd(data[cellid == c, ...], axis=0)
- mask[:, c, std == 0] |= BadPixels.DATA_STD_IS_ZERO.value
+ mask_std[:, c, std == 0] |= BadPixels.DATA_STD_IS_ZERO.value
- def correct_agipd(self, i_proc, first, last):
+ def offset_correction(self, i_proc: int, first: int, last: int):
"""
- Perform image-wise correction of data in shared memory
+ Perform image-wise offset correction for data in shared memory
:param first: Index of the first image to be corrected
:param last: Index of the last image to be corrected
:param i_proc: Index of shared memory array to process
"""
-
module_idx = self.shared_dict[i_proc]['moduleIdx'][0]
data = self.shared_dict[i_proc]['data'][first:last]
- bl_shift = self.shared_dict[i_proc]['blShift'][first:last]
rawgain = self.shared_dict[i_proc]['rawgain'][first:last]
gain = self.shared_dict[i_proc]['gain'][first:last]
cellid = self.shared_dict[i_proc]['cellId'][first:last]
- mask = self.shared_dict[i_proc]['mask'][first:last]
# first evaluate the gain into 0, 1, 2 --> high, medium, low
t0 = self.thresholds[module_idx][0]
t1 = self.thresholds[module_idx][1]
- rgain = None
- raw_data = None
+ # load raw_data and rgain to be used during gain_correction,
+ # if requested
if self.corr_bools.get('melt_snow'):
- rgain = rawgain / t0[cellid, ...]
- raw_data = np.copy(data)
-
+ self.shared_dict[i_proc]['t0_rgain'][first:last] = \
+ rawgain / t0[cellid, ...]
+ self.shared_dict[i_proc]['raw_data'][first:last] = np.copy(data)
+
# Often most pixels are in high-gain, so it's more efficient to
# set the whole output block to zero than select the right pixels.
gain[:] = 0
@@ -489,79 +491,121 @@ class AgipdCorrections:
off = calgs.gain_choose(gain, offsetb)
del offsetb
- tmask = self.mask[module_idx][:, cellid]
- msk = calgs.gain_choose_int(gain, tmask)
- del tmask
-
- # same for relative gain and then bad pixel mask
- if hasattr(self, "rel_gain"):
- # get the correct rel_gain depending on cell-id
- rc = self.rel_gain[module_idx][:, cellid]
- rel_cor = calgs.gain_choose(gain, rc)
- del rc
-
# subtract offset
data -= off
del off
+ def baseline_correction(self, i_proc:int, first:int, last:int):
+ """
+ Perform image-wise base-line shift correction for
+ data in shared memory via histogram or stripe
+
+ :param first: Index of the first image to be corrected
+ :param last: Index of the last image to be corrected
+ :param i_proc: Index of shared memory array to process
+ """
+
# before doing relative gain correction we need to evaluate any
# baseline shifts
# as they are effectively and additional offset in the data
- if (self.corr_bools.get('blc_noise') or
- self.corr_bools.get('blc_hmatch') or
- self.corr_bools.get('blc_stripes')):
-
- # do this image wise, as the shift is per image
- for i in range(data.shape[0]):
-
- # first correction requested may be to evaluate shift via
- # noise peak
- if self.corr_bools.get('blc_noise'):
- mn_noise = np.nanmean(self.noise[module_idx][0, cellid[i]]) #noqa
- dd, sh = baseline_correct_via_noise(data[i],
- mn_noise,
- gain[i],
- self.baseline_corr_noise_threshold) #noqa
- # if not we continue with initial data
- else:
- dd = data[i]
- sh = 0
-
- # if we have enough pixels in medium or low gain and
- # correction via hist matching is requested to this now
- gcrit = np.count_nonzero(gain[i] > 0) > 1000
- if (gcrit and self.corr_bools.get('blc_hmatch') and
- hasattr(self, "rel_gain")):
- dd2, sh2 = correct_baseline_via_hist(data[i],
- rel_cor[i],
- gain[i])
- data[i] = np.maximum(dd, dd2)
- sh = np.minimum(sh, sh2)
- # finally correct diagonal effects if requested
- if self.corr_bools.get('corr_asic_diag'):
- ii = data[i, ...]
- gg = gain[i, ...]
- adim = correct_baseline_via_hist_asic(ii, gg)
- data[i, ...] = adim
- # if there is not enough medium or low gain data to do an
- # evaluation, do nothing
- else:
- data[i, ...] = dd
- bl_shift[i] = sh
-
- if self.corr_bools.get('blc_stripes'):
- fmh = self.frac_high_med[module_idx][cellid[i]]
- dd, sh = baseline_correct_via_stripe(data[i, ...],
- gain[i, ...],
- msk[i, ...],
- fmh)
- data[i, ...] = dd
- bl_shift[i] = sh
+ if not any(self.blc_bools):
+ return
+
+ module_idx = self.shared_dict[i_proc]['moduleIdx'][0]
+ data = self.shared_dict[i_proc]['data'][first:last]
+ bl_shift = self.shared_dict[i_proc]['blShift'][first:last]
+ gain = self.shared_dict[i_proc]['gain'][first:last]
+ cellid = self.shared_dict[i_proc]['cellId'][first:last]
+ # output is saved in sharedmem to pass for correct_agipd()
+ # as this function takes about 3 seconds.
+ self.shared_dict[i_proc]['msk'][first:last] = \
+ calgs.gain_choose_int(gain,
+ self.mask[module_idx][:, cellid]) # noqa
+
+ if hasattr(self, "rel_gain"):
+ # Get the correct rel_gain depending on cell-id
+ self.shared_dict[i_proc]['rel_corr'][first:last] = \
+ calgs.gain_choose(gain,
+ self.rel_gain[module_idx][:, cellid]) # noqa
+
+ # do this image wise, as the shift is per image
+ for i in range(data.shape[0]):
+
+ # first correction requested may be to evaluate shift via
+ # noise peak
+ if self.corr_bools.get('blc_noise'):
+ mn_noise = np.nanmean(self.noise[module_idx][0, cellid[i]])
+ dd, sh = baseline_correct_via_noise(data[i], mn_noise,
+ gain[i],
+ self.baseline_corr_noise_threshold) # noqa
+ # if not we continue with initial data
+ else:
+ dd = data[i]
+ sh = 0
+
+ # if we have enough pixels in medium or low gain and
+ # correction via hist matching is requested to this now
+ gcrit = np.count_nonzero(gain[i] > 0) > 1000
+ if (gcrit and self.corr_bools.get('blc_hmatch') and
+ hasattr(self, "rel_gain")):
+ dd2, sh2 = correct_baseline_via_hist(data[i],
+ self.shared_dict[i_proc]['rel_corr'][first:last][i], # noqa
+ gain[i])
+ data[i] = np.maximum(dd, dd2)
+ sh = np.minimum(sh, sh2)
+ # finally correct diagonal effects if requested
+ if self.corr_bools.get('corr_asic_diag'):
+ ii = data[i, ...]
+ gg = gain[i, ...]
+ adim = correct_baseline_via_hist_asic(ii, gg)
+ data[i, ...] = adim
+ # if there is not enough medium or low gain data to do an
+ # evaluation, do nothing
+ else:
+ data[i, ...] = dd
+ bl_shift[i] = sh
+
+ if self.corr_bools.get('blc_stripes'):
+ fmh = self.frac_high_med[module_idx][cellid[i]]
+ dd, sh = baseline_correct_via_stripe(data[i, ...],
+ gain[i, ...],
+ self.shared_dict[i_proc]['msk'][first:last][i, ...], # noqa
+ fmh)
+ data[i, ...] = dd
+ bl_shift[i] = sh
+
+ def gain_correction(self, i_proc: int, first: int, last: int):
+ """
+ Perform several image-wise corrections for data in shared memory
+ e.g. Relative gain, FlatField xray correction, .....
+
+ :param first: Index of the first image to be corrected
+ :param last: Index of the last image to be corrected
+ :param i_proc: Index of shared memory array to process
+ """
+ module_idx = self.shared_dict[i_proc]['moduleIdx'][0]
+ data = self.shared_dict[i_proc]['data'][first:last]
+ gain = self.shared_dict[i_proc]['gain'][first:last]
+ cellid = self.shared_dict[i_proc]['cellId'][first:last]
+ mask = self.shared_dict[i_proc]['mask'][first:last]
+ rel_corr = self.shared_dict[i_proc]['rel_corr'][first:last]
+ msk = self.shared_dict[i_proc]['msk'][first:last]
+ # if baseline correction was not requested
+ # msk and rel_corr will still be empty shared_mem arrays
+ if not any(self.blc_bools):
+ msk = calgs.gain_choose_int(gain,
+ self.mask[module_idx][:, cellid])
+
+ # same for relative gain and then bad pixel mask
+ if hasattr(self, "rel_gain"):
+ # Get the correct rel_gain depending on cell-id
+ rel_corr = calgs.gain_choose(gain,
+ self.rel_gain[module_idx][:, cellid]) # noqa
# Correct for relative gain
if self.corr_bools.get("rel_gain") and hasattr(self, "rel_gain"):
- data *= rel_cor
- del rel_cor
+ data *= rel_corr
+ del rel_corr
# Adjust medium gain baseline to match highest high gain value
if self.corr_bools.get("adjust_mg_baseline"):
@@ -582,11 +626,13 @@ class AgipdCorrections:
if self.corr_bools.get("xray_corr"):
data /= self.xray_cor[module_idx]
+ # use sharedmem raw_data and t0_rgain
+ # after calculating it while offset correcting.
if self.corr_bools.get('melt_snow'):
- melt_snowy_pixels(raw_data, data, gain, rgain,
- self.snow_resolution)
- del raw_data
- del rgain
+ _ = melt_snowy_pixels(self.shared_dict[i_proc]['raw_data'][first:last], # noqa
+ data, gain,
+ self.shared_dict[i_proc]['t0_rgain'][first:last], # noqa
+ self.snow_resolution)
# Inner ASIC borders are matched to the same signal level
if self.corr_bools.get("match_asics"):
@@ -608,9 +654,13 @@ class AgipdCorrections:
del bidx
# Mask entire ADC if they are noise above a threshold
+ # TODO: Needs clarification if needed,
+ # the returned arg is not used.
if self.corr_bools.get("mask_noisy_adc"):
- make_noisy_adc_mask(msk, self.noisy_adc_threshold)
+ _ = make_noisy_adc_mask(msk,
+ self.noisy_adc_threshold)
+ # Copy the data across into the existing shared-memory array
mask[...] = msk[...]
def get_valid_image_idx(self, idx_base, infile, index_v=2):
@@ -935,7 +985,8 @@ class AgipdCorrections:
# add additional bad pixel information
if any(self.pc_bools):
- bppc = np.moveaxis(cons_data["BadPixelsPC"].astype(np.uint32), 0, 2) # noqa
+ bppc = np.moveaxis(cons_data["BadPixelsPC"].astype(np.uint32),
+ 0, 2)
bpixels |= bppc[..., :bpixels.shape[2], None]
slopesPC = cons_data["SlopesPC"].astype(np.float32)
@@ -1145,7 +1196,8 @@ class AgipdCorrections:
def allocate_images(self, shape, n_cores_files):
"""
- Allocate memory for image data
+ Allocate memory for image data and variable shared
+ between correction functions
:param shape: Shape of expected data (nImg, x, y)
:param n_cores_files: Number of files, handled in parallel
@@ -1164,7 +1216,16 @@ class AgipdCorrections:
self.shared_dict[i]['nImg'] = sharedmem.empty(1, dtype='i4')
self.shared_dict[i]['mask'] = sharedmem.empty(shape, dtype='u4')
self.shared_dict[i]['data'] = sharedmem.empty(shape, dtype='f4')
- self.shared_dict[i]['rawgain'] = sharedmem.empty(shape, dtype='u2')
+ self.shared_dict[i]['rawgain'] = sharedmem.empty(shape,
+ dtype='u2')
self.shared_dict[i]['gain'] = sharedmem.empty(shape, dtype='u1')
self.shared_dict[i]['blShift'] = sharedmem.empty(shape[0],
dtype='f4')
+ # Parameters shared between image-wise correction functions
+ self.shared_dict[i]['msk'] = sharedmem.empty(shape, dtype='i4')
+ self.shared_dict[i]['raw_data'] = sharedmem.empty(shape,
+ dtype='f4')
+ self.shared_dict[i]['rel_corr'] = sharedmem.empty(shape,
+ dtype='f4')
+ self.shared_dict[i]['t0_rgain'] = sharedmem.empty(shape,
+ dtype='u2')
diff --git a/notebooks/AGIPD/AGIPD_Correct_and_Verify.ipynb b/notebooks/AGIPD/AGIPD_Correct_and_Verify.ipynb
index e8e5e52265eca66111d76c8bd63109f0a2bc1c5e..2990da965082f29c46a4233c9febd3f9880755d8 100644
--- a/notebooks/AGIPD/AGIPD_Correct_and_Verify.ipynb
+++ b/notebooks/AGIPD/AGIPD_Correct_and_Verify.ipynb
@@ -358,9 +358,9 @@
" try:\n",
" gain_setting = get_gain_setting(control_fname, h5path_ctrl)\n",
" except Exception as e:\n",
- " print(f'Error while reading gain setting from: \\n{control_fname}')\n",
+ " print(f'ERROR: while reading gain setting from: \\n{control_fname}')\n",
" print(e)\n",
- " print(\"Set gain settion to 0\")\n",
+ " print(\"Set gain setting to 0\")\n",
" gain_setting = 0\n",
" "
]
@@ -516,24 +516,34 @@
" step_timer.start()\n",
" \n",
" img_counts = pool.starmap(agipd_corr.read_file, enumerate(file_batch))\n",
- " step_timer.done_step('Load')\n",
+ " step_timer.done_step('Loading data from files')\n",
" \n",
" # Evaluate zero-data-std mask\n",
" pool.starmap(agipd_corr.mask_zero_std, itertools.product(\n",
" range(len(file_batch)), np.array_split(np.arange(agipd_corr.max_cells), n_cores_correct)\n",
" ))\n",
" step_timer.done_step('Mask 0 std')\n",
+ "\n",
+ " # Perform offset image-wise correction\n",
+ " pool.starmap(agipd_corr.offset_correction, imagewise_chunks(img_counts))\n",
+ " step_timer.done_step(\"Offset correction\")\n",
" \n",
- " # Perform image-wise correction\n",
- " pool.starmap(agipd_corr.correct_agipd, imagewise_chunks(img_counts))\n",
- " step_timer.done_step(\"Image-wise correction\")\n",
" \n",
- " # Perform cross-file correction parallel over asics\n",
- " pool.starmap(agipd_corr.cm_correction, itertools.product(\n",
- " range(len(file_batch)), range(16) # 16 ASICs per module\n",
- " ))\n",
- " step_timer.done_step(\"Common-mode correction\")\n",
+ " if blc_noise or blc_stripes or blc_hmatch:\n",
+ " # Perform image-wise correction\n",
+ " pool.starmap(agipd_corr.baseline_correction, imagewise_chunks(img_counts))\n",
+ " step_timer.done_step(\"Base-line shift correction\")\n",
" \n",
+ " if common_mode:\n",
+ " # Perform cross-file correction parallel over asics\n",
+ " pool.starmap(agipd_corr.cm_correction, itertools.product(\n",
+ " range(len(file_batch)), range(16) # 16 ASICs per module\n",
+ " ))\n",
+ " step_timer.done_step(\"Common-mode correction\")\n",
+ " \n",
+ " # Perform image-wise correction\n",
+ " pool.starmap(agipd_corr.gain_correction, imagewise_chunks(img_counts))\n",
+ " step_timer.done_step(\"Image-wise correction\")\n",
" \n",
" # Save corrected data\n",
" pool.starmap(agipd_corr.write_file, [\n",
diff --git a/notebooks/AGIPD/AGIPD_Retrieve_Constants_Precorrection.ipynb b/notebooks/AGIPD/AGIPD_Retrieve_Constants_Precorrection.ipynb
index 87bb034efbd1cb504fb6cf6a4b407dd38a168d2d..e19a03b99b81b8394b893ed27e44288fabe9e932 100644
--- a/notebooks/AGIPD/AGIPD_Retrieve_Constants_Precorrection.ipynb
+++ b/notebooks/AGIPD/AGIPD_Retrieve_Constants_Precorrection.ipynb
@@ -167,8 +167,11 @@
" try:\n",
" gain_setting = get_gain_setting(control_fname, h5path_ctrl)\n",
" except Exception as e:\n",
- " print(f'Error while reading gain setting: {e}\\n')\n",
- " \n",
+ " print(f'ERROR: while reading gain setting from: \\n{control_fname}')\n",
+ " print(e)\n",
+ " print(\"Set gain setting to 0\")\n",
+ " gain_setting = 0\n",
+ "\n",
"print(f\"Gain setting: {gain_setting}\")\n",
"print(f\"Detector in use is {karabo_id}\")\n",
"\n",
diff --git a/requirements.txt b/requirements.txt
index 039431d2c1091b5c0f5908c09601473bdebc43c3..7b0af0733e86e8c349a36a14fe291d1a209b482a 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -26,6 +26,7 @@ metadata_client == 3.0.5
nbclient == 0.5.1
nbconvert == 5.6.1
nbformat == 5.0.7
+notebook == 6.1.5
numpy == 1.19.1
prettytable == 0.7.2
princess == 0.2