Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://www.mrao.cam.ac.uk/~bn204/almasim08/papers/OnLineSimulation.pdf Äàòà èçìåíåíèÿ: Tue Dec 2 15:00:19 2008 Äàòà èíäåêñèðîâàíèÿ: Tue Oct 2 13:07:04 2012 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: ion drive

On line simulation in ALMA Software

Robert Lucas

ALMA Simulation workshop


Summary

· · · · · ·

Need for hardware simulation for software testing Requirements for on-line simulation A Shared Simulator Limitations Current capabilities Further work

ALMA Simulation workshop

2


Need for on-line simulation
· Main motivation is to test software
­ At the subsystem level: simulate the hardware for which specific software (drivers) are written or other pieces of software · Not the subject of this talk ­ At the system level: use all the software, without any hardware · This is done to integrate all the subsystems together (integrated tests). · This is mainly done in the process of software integration. · We have multiple test systems (STEs) to test integrated software on the different development sites (ESO, NRAO, NAOJ, ALMA/Chile)
ALMA Simulation workshop 3


Requirements for on-line simulation
· Perform an end-to-end simulation:
­ Put Scheduling Blocks (SB) in the Archive using the Observing Tool (OT) ­ Drive the system from the Operator interface (OMC) ­ Schedule a SB ­ Execute it, including On-line Calibration and Quick-Look ­ Check the resulting data (ASDM) in the Archive

· To do this we need to:
ALMA Simulation workshop 4 ­ Simulate all hardware devices by their own simulators (e g antenna mounts local oscillators


Shared Simulator
· In order to get meaningful data, we need, e.g. :

· In principle do this fast enough to simulate the actual observing process
­ In practice do this on a timescale of 48ms (1 TE).

­ The total power detectors know what the antenna is doing (on-source or off-source?) ­ The correlator knows what the interferometry specific hardware is doing (delay and fringe tracking).

· We have a centralized software component (the Shared Simulator)

­ This component communicates with the individual hardware simulators so that those that produce data do this in a consistent way.
ALMA Simulation workshop 5


Example 1
· To calculate the total power output of a detector, the shared simulator would:
­ Get the current antenna pointing direction · Talk to mount simulator ­ Get the actual sources in the beam at current time · Have a model of the sky ­ Compute the expected detector output · Know the current attenuator setup · Know the LO1, LO2 frequencies for that baseband · Know to which receiver band that baseband is connected (USB, LSB) · Know the receiver gain and noise at this frequency · Know the atmosphere properties at this frequency
ALMA Simulation workshop 6


Example 2
· To calculate the correlator output in a baseband/baseline (ab), the shared simulator would need to:

­ Get the current antenna pointing directions for antennas a and b · Talk to mount simulators, focus control simulators ­ Get the actual sources in the beam at current time · Have a model of the sky · Compute the visibilities for sources in the primary beam ­ Compute the expected correlation function for baseline ab · Know the current attenuator setup · Know the LO1, LO2 frequencies for that baseband, and the phase commands and phase rates applied to those LOs · Know the delays applied at various stages (sampler clock, correlator) · Know to which receiver band that baseband is connected (USB, LSB) · Know the receiver gain and noise at this frequency · Know the atmosphere properties (absorption, added path) at this frequency

ALMA Simulation workshop

7


Limitations
· This can become very complex, given the complexity of ALMA hardware. · All devices in the data path can have an effect on the correlator output · These effects can be very subtle (there might even be some that we · Sending very often a full set of correlation functions to the correlator simulators can be very heavy in terms of data rate... · Thus we needed to simplify a lot; fortunately the scope of the simulation is limited. · To start with, some of the control software was not yet developed (e.g. LO control), so we needed to bypass those.
do not anticipate fully at this time, hopefully a few...)

ALMA Simulation workshop

8


Actual scope?
· Provide a simulation realistic enough that we could read the data through data reduction software, and see that it is not grossly wrong · The goal is to check that the software is:
­ Not crashing ­ Going to the end of the Scheduling Block ­ Provide the expected values in the various columns of the ASDM tables ­ Even more: we see the source, we see fringes, the phase is not drifting...

· There more efficient ways of doing data simulations that are capable of producing results suitable to fully qualify the observing strategies and data reduction algorithms of ALMA .
­ That's of course the rest of this workshop.
ALMA Simulation workshop 9


Current simulation capabilities
· Simulate 2 point sources in the sky (e.g. 1 object, 1 calibrator), with flat spectrum · Assume rectangular bandpasses · Flat-spectrum system temperatures · No atmospheric fluctuations · Describe correlation functions by a few parameters (amplitude, phase, delay, phase rate, gaussian noise level), that the correlator simulator uses to generate the lags from a simple formula, including random Gaussian noise.
ALMA Simulation workshop 10


Other functionality
· Switch on/off the use of `actual' antenna pointing directions (assume perfect pointing, focus tracking) · Switch on/off use of LO tracking, delay tracking (assume it works, i.e. replace by ideal values) · Actually most if not all of the simulations we've made in the last few years have been using these `off' modes...

ALMA Simulation workshop

11


Further work?
· May be justified by the unavailability of a working ALMA interferometer after ATF closes, and until OSF interferometry is achieved. · May also be justified by the improved simulation capabilities in Control Software (simulating more than 2 antennas in the same computer) which means that we may need only two computers to run such a simulation (Simulation FBT) · Use full implemented functionality · Possibilities: add WVR receivers, simulating e.g. the pathlength fluctuations by a simple waveform. · Having a casa script to check the final product of a simulation run would be useful.

ALMA Simulation workshop

12