Page History
...
- 2018-11-18-epix10ka2m-metrology-map.pdf
- 2018-11-18-epix10ka2m-metrology-map.pptx
- 2018-11-15-Metrology-epix10ka2m.xlsx
Front and back side of the new detector epix10ka2m.1 from mfxc00118 2020-07-dd
2020-10-04 MEC epix10kaquad 0 at optical metrology
2020-10-06 MEC epix10kaquad 0 front and back
2021-04-10 UED epix10kaquad
Comments on orientation of epix10ka2m parts from Matt
...
Gain factors from charge injection default and measured
gain | charge injection | current default | measured (ADU / keV) | 2020-08-03 Gabriel (ADU / keV) - use as default |
---|---|---|---|---|
L | 0.46 | 0.01 | 0.139 | 0.164 |
M | 15. | 0.3(3) | 4.5 | 5.466 |
H | 46.7 | 1 | 13.9 | 16.40 |
Gain factors default vs charge injection
...
gain from charge injection:
constants | Mean | RMS | RMS / MEAN |
---|---|---|---|
default | 1117.7 | 62.72 | 0.05618 |
charge injection | 1177.5 | 66.79 | 0.05672 |
Conclusion: in this test charge injection gainci constants do not improve gain factors comparing to default
Test of the gain switching modes
offset calibration: exp=xcsx35617:run=544; its timestamp 20181129124822 faked for earlier dark calibrations by reference from 20180101000000
dark runs: 413, 416, 417, 420 of xcsx35617
gain factors M, H=1, L= 0.2, 0.25, 0.3, 0.33333, 0.4
gain map images show that lateral and central-most pixels in mode H, M, "water ring" region pixels switched to L
data:
- AML: exp=xcsx35617:run=419, event 3
- AHL: exp=xcsx35617:run=414, event 3
Masks
mask_geo
mask_geo = det.mask_geo(par, mbits=3, width=10, wcentral=5)
- mbits = 1 - masks edges, +2 - masks central rows and columns.
- width - number of edge rows or columns to mask, def=1
- wcentral - number of central rows or columns to mask, def=1
plot for mask_geo + 1:
status_as_mask
- use pixel_status for exp=xcsx35617:run=544
- mask_status = det.status_as_mask(par, mode=0, indexes=(0,1,2,3,4))
- mode 0/1/2 masks zero/four/eight neighbors around each bad pixel
- indexes=(0,1,2,3,4) # indexes stand for FH, FM, FL, AHL-H, AML-M, respectively. Derived modes have the same status arrays.
found number of bad pixels
- 2802 for F gain modes and
- 3253 for all F + A mode
plots for mask_status + 1 for mode=0, 1 and 2:
Combined mask
mask = det.mask(par, calib=False, status=True, edges=True, central=True, width=10, wcentral=5, mode=0)
Calibrated data and mask
Image and spectrum for
- nda = calib_epix10ka_any(det, evt)
- nda *= det.mask(par, calib=False, status=True, edges=True, central=True, width=1, wcentral=1, mode=0)
Manual alignment on 2019-05-06
Data
Ring-data (npy) arrays were provided for xcsx35617 run 400 by Silke, available under
- /reg/g/psdm/detector/alignment/epix10ka2m/calib-mfx-epix10ka2m-01-2018-11-15/2019-05-06-geometry-alignment/
Alignment tool
Manual Detector alignment tool (geo) is used for alignment. There is no automated geometry optimization in this tool.
Initial geometry
Alignment is started with the best geometry file obtained after optical metrology measurements for two quads, like
/reg/g/psdm/detector/alignment/epix10ka2m/calib/Epix10ka2M::CalibV1/NoDetector.0:Epix10ka2M.0/geometry/geo-epix10ka2m-v180
or
/reg/d/psdm/xcs/xcsx35617/calib/Epix10ka2M::CalibV1/XcsEndstation.0:Epix10ka2M.0/geometry/398-398.data
Alignment procedure
Quads' x0,y0 - center positions ONLY have been tuned as explained here:
1) Q0 and Q1 were moved together relative to the image center, because their geometry is constrained from optical metrology.
2) then Q2 and Q3 were moved independently in order to get consistent "to my eye" image relative to a set of drown circles.
Geometry for panels inside Q2 and Q3 is set from design geometry, and I do not feel that could do better job moving panels in quad.
There are some regular alignment issues with this detector; if I tune nicely (with precision ~ pixel size) rings in the middle of radial range,
then internal and external rings may be misaligned. This may be due to small tilt of the detector or non-accounted z position of panels
w/o optical metrology.
Results
Resulting geometry for this data looks like on attached image.
All files are available under
- /reg/g/psdm/detector/alignment/epix10ka2m/calib-mfx-epix10ka2m-01-2018-11-15/2019-05-06-geometry-alignment/
Recommendation for further geometry improvement
The only reliable procedure to get correct detector geometry is an 3-d optical metrology of entire detector.
After that one would need to adjust precisely
1) detector center relative to image with rings
2) sample-to-detector distance
3) detector plane tilts.
2020-06-25 New Optical Metrology
Code Block | ||||
---|---|---|---|---|
| ||||
Hi Philip, Chris, and Mikhail,
My apologies for not getting this to you sooner.
Additional apologies in that the order we took the measurements in
is rotated from our "standard" and we were unable to set the origin
to zero. In the picture the XPP mounting bars on on the left side and
so this would be UP.
A map of the sensor corners measurement ordering and the quad
number locations is attached along with a picture of the setup. The
origin (not 0-0-0) is in the upper left of the picture.
Along with the Excel spreadsheet.
Please ask questions as I know this can be confusing.
Thanks,
Chris |
References
Default gain correction factors
Code Block | ||||
---|---|---|---|---|
| ||||
Blaj, Gabriel <blaj@slac.stanford.edu> Mon 8/3/2020 6:52 PM
To: Hart, Philip Adam;
Dragone, Angelo;
Kenney, Christopher J.;
Dubrovin, Mikhail;
O'Grady, Paul Christopher;
Hansson, Conny;
McKelvey, Mark E
Hi, Here are some good starting values for the ADC to keV conversion:
High gain: 132 ADU / 8.05 keV = 16.40 ADU/keV
Medium gain: 132 ADU / 8.05 keV / 3 = 5.466 ADU/keV
Low gain: 132 ADU / 8.05 keV / 100 = 0.164 ADU/keV
Of course, a gain calibration is preferable.
The same numbers work in both fixed and auto-ranging gain modes.
Thanks,
Gabriel
=========
Blaj, Gabriel <blaj@slac.stanford.edu> Mon 8/3/2020 7:13 PM
Hi, For the long integration time, I don't have a set of magic numbers, but this iterative procedure should yield optimal settings:
Cool the camera as low as possible, just a few degrees over the minimum temperature to allow temperature stabilization by the PID control loop (either the chiller PID for the large cameras, or the Peltier PID in the small cameras). Of course, the small cameras can be cooled much lower than the large ones.
Start with the default LCLS settings (I believe both AsicAcqWidth and R0toAcq are set to 100us by default)
0 AsicAcqWidth should be optimized for the experiment. With a very cold camera (e.g., < -15ºC) you could go to 5ms. A good starting value would be 1ms.
1 Set AsicAcqWidth to, e.g., 1 ms
2 Set R0toACQ time to 100us
3 Decrease frame rate until no frames are dropped
4 Set the X-ray source to a low flux (0.01-0.05 photons/pixel/frame?)
5 Try to get a uniform illumination
6 Repeat:
- Calibrate dark
- Take many frames and integrate them
- Look if the resulting image is uniform or has a strange sawtooth pattern over each ASIC
- If no, try reducing R0toACQ
- If yes, try increasing R0toAACQ
- Increase/decrease frame rate to the maximum frame rate that runs reliably (no dropped frames).
6 Until an optimum is found.
For an idea how the strange sawtooth pattern looks, you could try setting:
AsicAcqWidth = 1ms
R0toAcq = 50us, or 20us.
Thanks,
Gabriel |
Test of the gain switching modes
offset calibration: exp=xcsx35617:run=544; its timestamp 20181129124822 faked for earlier dark calibrations by reference from 20180101000000
dark runs: 413, 416, 417, 420 of xcsx35617
gain factors M, H=1, L= 0.2, 0.25, 0.3, 0.33333, 0.4
gain map images show that lateral and central-most pixels in mode H, M, "water ring" region pixels switched to L
data:
- AML: exp=xcsx35617:run=419, event 3
- AHL: exp=xcsx35617:run=414, event 3
Masks
mask_geo
mask_geo = det.mask_geo(par, mbits=3, width=10, wcentral=5)
- mbits = 1 - masks edges, +2 - masks central rows and columns.
- width - number of edge rows or columns to mask, def=1
- wcentral - number of central rows or columns to mask, def=1
plot for mask_geo + 1:
status_as_mask
- use pixel_status for exp=xcsx35617:run=544
- mask_status = det.status_as_mask(par, mode=0, indexes=(0,1,2,3,4))
- mode 0/1/2 masks zero/four/eight neighbors around each bad pixel
- indexes=(0,1,2,3,4) # indexes stand for FH, FM, FL, AHL-H, AML-M, respectively. Derived modes have the same status arrays.
found number of bad pixels
- 2802 for F gain modes and
- 3253 for all F + A mode
plots for mask_status + 1 for mode=0, 1 and 2:
Combined mask
mask = det.mask(par, calib=False, status=True, edges=True, central=True, width=10, wcentral=5, mode=0)
Calibrated data and mask
Image and spectrum for
- nda = calib_epix10ka_any(det, evt)
- nda *= det.mask(par, calib=False, status=True, edges=True, central=True, width=1, wcentral=1, mode=0)
Manual alignment on 2019-05-06
Data
Ring-data (npy) arrays were provided for xcsx35617 run 400 by Silke, available under
- /reg/g/psdm/detector/alignment/epix10ka2m/calib-mfx-epix10ka2m-01-2018-11-15/2019-05-06-geometry-alignment/
Alignment tool
Manual Detector alignment tool (geo) is used for alignment. There is no automated geometry optimization in this tool.
Initial geometry
Alignment is started with the best geometry file obtained after optical metrology measurements for two quads, like
/reg/g/psdm/detector/alignment/epix10ka2m/calib/Epix10ka2M::CalibV1/NoDetector.0:Epix10ka2M.0/geometry/geo-epix10ka2m-v180
or
/reg/d/psdm/xcs/xcsx35617/calib/Epix10ka2M::CalibV1/XcsEndstation.0:Epix10ka2M.0/geometry/398-398.data
Alignment procedure
Quads' x0,y0 - center positions ONLY have been tuned as explained here:
1) Q0 and Q1 were moved together relative to the image center, because their geometry is constrained from optical metrology.
2) then Q2 and Q3 were moved independently in order to get consistent "to my eye" image relative to a set of drown circles.
Geometry for panels inside Q2 and Q3 is set from design geometry, and I do not feel that could do better job moving panels in quad.
There are some regular alignment issues with this detector; if I tune nicely (with precision ~ pixel size) rings in the middle of radial range,
then internal and external rings may be misaligned. This may be due to small tilt of the detector or non-accounted z position of panels
w/o optical metrology.
Results
Resulting geometry for this data looks like on attached image.
All files are available under
- /reg/g/psdm/detector/alignment/epix10ka2m/calib-mfx-epix10ka2m-01-2018-11-15/2019-05-06-geometry-alignment/
Recommendation for further geometry improvement
The only reliable procedure to get correct detector geometry is an 3-d optical metrology of entire detector.
After that one would need to adjust precisely
1) detector center relative to image with rings
2) sample-to-detector distance
3) detector plane tilts.
References
- EPIX10KA2M References
- EPIX10KA
- EPIX10KA2M data images and geometry
- Production of calibration constants for multi-panel epix10ka detectors
- EPIX10KA2M Charge injection fit issue
- EPIX10KA2M issue with panel 6 in mfxc00318
- EPIX10KA2M issue with panel 12 in xcsc00118 2M.0
- Optical metrology of epix10ka2m.1 from 2020-02-25
- EPIX10KA panel flatness measurement
- Jungfrau and Epix10ka Calibration
- Detector alignment tool
- 2020-08-20-new-x-ray-detector-snaps-1000-atomic-level-pictures-second-natures-ultrafast
- EPIX10KA
- Jungfrau and Epix10ka Calibration
- Detector alignment tool
- /reg/g/psdm/detector/alignment/epix10ka2m/calib-mfx-epix10ka2m-01-2018-11-15