[TRCL-570][feat] introduce startScan Event to synchronize CCD with e-beam during SPARC acq #3027

pieleric · 2025-02-06T13:26:07Z

From the moment that the acquisition ask the e-beam scanner to start and the e-beam is actually at the right position, it takes some time. It's in the order of a few ms. So far, the only e-beam scanner was semcomedi, and it was (surprisingly) constant in terms of time. It would take between 2 and 5 ms. So it was good enough to just always wait 5ms.

We now also have semnidaq, and the NI-DAQ library introduces a lot longer latency. While it can be as "fast" as 10ms, it sometimes takes 40ms to start a scan. There are probably ways to make it a little faster, but it's unlikely it will be possible to always keep all scanners < 5ms.

=> Use a new Event, startScan, for the e-beam scanner to report the e-beam is ready. Then, we can use it to directly start a CCD acquisition.

No code ever used this functionality. The semnidaq has something a little similar, but it's a hardware event. So it actually could even brinkg confusion. Now that we'll add another Event, remove this code.

Introduce a new (and almost first) software Event on the e-beam scanners: startScan. It is emitted at the very beginning of the first frame of a scan. If a scan acquires multiple frames, startScan is emitted just once. This will be used for better software synchronization of SPARC acquisitions.

… e-beam From the momement that the acquisition ask the e-beam scanner to start and the e-beam is actually at the right position, it takes *some* time. It's in the order of a few ms. So far, the only e-beam scanner was semcomedi, and it was (surprisingly) constant in terms of time. It would take between 2 and 5 ms. So it was good enough to just always wait 5ms. We now also have semnidaq, and the NI-DAQ library introduces a lot longer latency. While it can be as "fast" as 10ms, it sometimes takes 40ms to start a scan. There are probably ways to make it a little faster, but it's unlikely it will be possible to always keep all scanners < 5ms. => Use a new Event, startScan, for the e-beam scanner to report the e-beam is ready. Then, we can use it to directly start a CCD acquisition.

K4rishma · 2025-02-17T14:29:41Z

src/odemis/driver/test/semnidaq_test.py

+        self.assertEqual(evt_counter.count, 1)
+        self.scanner.startScan.unsubscribe(evt_counter)
+
+


to remove the extra space between the functions

tmoerkerken · 2025-02-17T14:47:42Z

src/odemis/acq/stream/_sync.py

@@ -1887,26 +1888,10 @@ def _acquireImage(self, n, px_idx, img_time, sem_time,
                raise CancelledError()

            # subscribe to _subscribers
+            # As soon as the e-beam start scanning (which can take a couple of ms), the


Suggested change

# As soon as the e-beam start scanning (which can take a couple of ms), the

# As soon as the e-beam starts scanning (which can take a couple of ms), the

tmoerkerken · 2025-02-17T14:49:09Z

src/odemis/acq/stream/_sync.py

@@ -1887,26 +1888,10 @@ def _acquireImage(self, n, px_idx, img_time, sem_time,
                raise CancelledError()

            # subscribe to _subscribers
+            # As soon as the e-beam start scanning (which can take a couple of ms), the
+            # startScan event is sent, which triggers the acquisition of one CCD frame.
            for s, sub in zip(self._streams[:-1], self._subscribers[:-1]):


Can you comment on why the last element is excluded?

tmoerkerken · 2025-02-17T14:50:56Z

src/odemis/acq/stream/_sync.py

@@ -2119,12 +2112,11 @@ def _runAcquisitionScanStage(self, future):
                    if self._acq_state == CANCELLED:
                        raise CancelledError()

+                    # Start e-beam scan. As soon as it really start, a startScan event is sent, which


Suggested change

# Start e-beam scan. As soon as it really start, a startScan event is sent, which

# Start e-beam scan. As soon as it really starts, a startScan event is sent, which

tmoerkerken · 2025-02-17T14:51:28Z

src/odemis/acq/stream/_sync.py

@@ -2008,9 +2001,9 @@ def _runAcquisitionScanStage(self, future):
        # The idea of the acquiring with a scan stage:


Suggested change

# The idea of the acquiring with a scan stage:

# The idea of acquisition with a scan stage:

tmoerkerken · 2025-02-17T14:55:03Z

src/odemis/driver/semnidaq.py

@@ -2251,6 +2253,23 @@ def _set_state(self, active):
                time.sleep(self._activation_delay)


+class EventOnce(model.Event):


This is duplication of the function at the semcomedi driver. Maybe put it in a shared module?

Indeed. moved to driver.EventOnce

K4rishma · 2025-02-17T14:40:35Z

src/odemis/driver/simsem.py

@@ -540,6 +544,9 @@ def _acquire_thread(self, callback):
                # as in Odemis the convention for SEM is that the ebeam waits
                # for _all_ the detectors to be ready before scanning.
                self.data._waitSync()
+                if first_frame:
+                    self.parent._scanner.startScan.notify()
+                    first_frame = False


Does this automatically stop notifying something which is similar to the unsubscribe used in the testcases when self._acquisition_must_stop.clear() is called in finally: section?

K4rishma · 2025-02-17T14:58:23Z

src/odemis/acq/stream/_sync.py

+                                # Temporarily switch the CCD to a different event trigger, so that it
+                                # doesn't get triggered while the leech is running (because it could use the
+                                # e-beam, which would send a startScan event)
+                                self._ccd_df.synchronizedOn(self._ccd.softwareTrigger)


Maybe an assert line can be used to check that the notification boolean value of startscan is False before reusing the real trigger? Same for _acquireImage

pieleric added 4 commits February 6, 2025 14:26

[clean-up] driver semcomedi: remove newPixel event support

2b1dfed

No code ever used this functionality. The semnidaq has something a little similar, but it's a hardware event. So it actually could even brinkg confusion. Now that we'll add another Event, remove this code.

[doc] stream sync improve comments and logs

5822427

pieleric requested review from tmoerkerken and K4rishma February 6, 2025 13:26

pieleric changed the title ~~[feat] introduce startScan Event to synchronize CCD with e-beam during SPARC acq~~ [TRCL-570][feat] introduce startScan Event to synchronize CCD with e-beam during SPARC acq Feb 6, 2025

K4rishma reviewed Feb 17, 2025

View reviewed changes

tmoerkerken requested changes Feb 17, 2025

View reviewed changes

K4rishma approved these changes Feb 17, 2025

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[TRCL-570][feat] introduce startScan Event to synchronize CCD with e-beam during SPARC acq #3027

[TRCL-570][feat] introduce startScan Event to synchronize CCD with e-beam during SPARC acq #3027

pieleric commented Feb 6, 2025 •

edited

Loading

K4rishma Feb 17, 2025

tmoerkerken Feb 17, 2025

tmoerkerken Feb 17, 2025

tmoerkerken Feb 17, 2025

tmoerkerken Feb 17, 2025

tmoerkerken Feb 17, 2025

pieleric Feb 17, 2025

K4rishma Feb 17, 2025

K4rishma Feb 17, 2025

		self.assertEqual(evt_counter.count, 1)
		self.scanner.startScan.unsubscribe(evt_counter)

	# As soon as the e-beam start scanning (which can take a couple of ms), the
	# As soon as the e-beam starts scanning (which can take a couple of ms), the

	# Start e-beam scan. As soon as it really start, a startScan event is sent, which
	# Start e-beam scan. As soon as it really starts, a startScan event is sent, which

		@@ -2008,9 +2001,9 @@ def _runAcquisitionScanStage(self, future):
		# The idea of the acquiring with a scan stage:

	# The idea of the acquiring with a scan stage:
	# The idea of acquisition with a scan stage:

		@@ -2251,6 +2253,23 @@ def _set_state(self, active):
		time.sleep(self._activation_delay)


		class EventOnce(model.Event):

[TRCL-570][feat] introduce startScan Event to synchronize CCD with e-beam during SPARC acq #3027

Are you sure you want to change the base?

[TRCL-570][feat] introduce startScan Event to synchronize CCD with e-beam during SPARC acq #3027

Conversation

pieleric commented Feb 6, 2025 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

pieleric commented Feb 6, 2025 •

edited

Loading