Add HWP Sequencer app

[mileslucas]

HWP sequencer manages the state machine for the polarimeter, which moves the HWP stage and toggles data acquistion on and off.

To be tested.

[mileslucas]

Testing works fine using a non-threaded blocking appLogic method with a magic sleep timer. Need to move to a threaded call to moveHwp which should wait until the HWP is in the desired position.

[mileslucas]

This is tested working on exao1

• acquisition time is greater than or equal to the timePerPos property
• the hwp moves to correct position and blocks until that position is reached
• the hwp positions work regardless of hwp adi tracking being turned on
• setting numCyles to -1 correctly loops forever
• setting numCycles to >0 correctly ends after the number of loops
• the lastCycle toggle appropriately stops sequencing at the final hwp position
• toggle the sequence switch immediately ends the acquistion
• if the hwpsequence app crashes/stops it will stop saving data
• No blocking code is inside appLogic, it's inside a separate thread

[mileslucas]

Actually, one question to resolve is whether we should be triggering obs_on or triggering the fxngensync output.

Triggering obs_on is noisy on maggieo_x and leads to an explosion of folders, so it might be better to turn the trigger off instead, but this conflicts with the lab rules I just wrote for #228

[mileslucas]

New behavior using fxngensync output to pause acquisition during HWP movement is tested and working

[mileslucas]

Final piece of logic to implement is some kind of block to wait for the current frame to readout before moving HWP--I think I can just wait for a semaphore from the SHM for one of the camscis.

[mileslucas]

I've moved the app to run on ICC so the SHMs are as fresh as possible. The semaphore waits are tested to work with both fast exposure times (5 Hz) and slow exposure times (10s). Long exposures have little to no observing efficiency loss because we turn the external trigger off and then wait for the semaphore, which means there are no intermediate exposures that lock us up for a full exposure time before continuing saving.

I think this is ready to go now!

[mileslucas]

depends on #244 now but otherwise should be ready to go

[mileslucas]

new features as of 2025/12/09

• If data is already being saved, turn it off via obs_on
• Make sure the fxngensync output is turned on when appShutdown is called to keep cameras running even if this app breaks
• Move HWP back to 0 deg after sequence is finished
• Update to using shmimMonitor which means the semaphore wait still works even when camera ROI changes or any other case where the shmim changes

[mileslucas]

@jaredmales this is the block of code that is screwed up--the sympton is that regardlesss of camera readout time, on the first iteration of the sequence the sem_wait call is instantaneous. only on the first iteration. This is obvious when using long exposures, if you set EXPTIME to 10 seconds and start a sequence, the obs_on toggle will turn on way before the current frame is read out. Subsequent sem_waits do exactly what's expected--blocks until a frame is received and the processImage method is called.

It is crazy bizarre that the same exact code has different behavior. The startSequencing method does not touch the semaphore, nor does appLogic. One of the only things that's different is the m_startSaving flag (true on first iteration, false otherwise), and the indices for the current cycle status.

MagAOX/apps/hwpSequencer/hwpSequencer.hpp

Lines 434 to 469 in 6e99490

	// DEBUG
	int semval;
	sem_getvalue( &m_smSemaphore, &semval);
	std::cerr << "SEMVALUE before flush: " << semval << std::endl;
	XWC_SEM_FLUSH(m_smSemaphore);
	// DEBUG
	sem_getvalue( &m_smSemaphore, &semval);
	std::cerr << "SEMVALUE after flush: " << semval << std::endl;
	// Stop triggering
	m_indiP_fxngenOutput["value"] = "Off";
	sendNewProperty(m_indiP_fxngenOutput);
	sem_getvalue( &m_smSemaphore, &semval);
	std::cerr << "SEMVALUE after fxngenoutput changed: " << semval << std::endl;
	while (m_fxngenOutp)
	mx::sys::milliSleep(m_hwpWait);
	sem_getvalue( &m_smSemaphore, &semval);
	std::cerr << "SEMVALUE after fxngenoutput confirmed: " << semval << std::endl;
	sem_getvalue( &m_smSemaphore, &semval);
	std::cerr << "SEMVALUE before wait: " << semval << std::endl;
	// Wait for current frame to arrive
	if (sem_wait(&m_smSemaphore) < 0)
	{
	log<software_error>({ __FILE__, __LINE__, "failed waiting for semaphore" });
	return -1;
	}
	sem_getvalue( &m_smSemaphore, &semval);
	std::cerr << "SEMVALUE after wait: " << semval << std::endl;

[jaredmales]

the sleep function takes an unsigned as argument, so it's integer seconds. You probably want to use mx::sys::microSleep(m_timePerPos*1e6);

[jaredmales]

see dfd2f8

[jaredmales]

same as above. In this case it's not sleeping at all.

[jaredmales]

see dfd2f8