DEVELOPMENT AND TEST FACTORS

Performances

The ATCS system is based on the VLT Unit Telescope Control system and is deployed on a similar hardware.

For this reason no performance problems are expected.

The only critical component is the Sensor LCU. This LCU will be responsible for many critical real-time subsystems (Field Stabilization Sensor, Field Acquisition Sensor, Chopping and FSS Translation Stage tracking system) in a combination never tried on the Unit Telescopes. Although using a PowerPC CPU much more powerful than the reference 68k CPUs, we could have performance problems in particular in terms of synchronization among the various control loops.

This potential problem will be assessed by the development of the foreseen prototype.

In case the CPU is not able to provide the required performances, a backup solution will be to split the LCU in two units. The Field Acquisition Sensor and, eventually, the FSS Translation Stage could be easily moved on a second CPU. In order to save costs and if necessary, this second CPU could be a 68k "recycled" from the Unit Telescopes, where all major LCUs are now being replaced by PowerPC.

In the following paragraphs you will find the results of the performance analysis concerning the needed computer power for the ATCS.

As a starting point the current loads of the TCS on the UT machines are taken. The estimations for the ATCS machines are based on these numbers.

TCS (UTs) analysis:

The following table shows the CPU loads and memory usage of TCS used on the UTs:

Machine	Descr.	Type	CPU load	Memory usage (MB)	Mode
wt0tcs	telescope control	WS-JJ240	23%	N/A	N/A
wa0tcs	guide/acquisition	WS-C110	64%	N/A	N/A
la0agc	guiding LCU FS 30Hz, 15x15 window	LCU-68k	74%	6.4	real
la0iac	image analysis LCU 10s cycle	LCU-68k	60%	12.3	real
lt0alt	altitude axis LCU tracking + guiding	LCU-68k	50%	5	real
lt0az	azimuth axis LCU tracking + guiding	LCU-68k	50%	5	real
lt0enc	enclosure LCU	LCU-68k	60%	5.6	sim
lt0hb	hydrostatic bearings	LCU-68k	10%	4.6	sim
lt0m2	m2 control	LCU-68k	75%	6.2	sim
lr0m2	m2 rapid guiding	LCU-68k	14%	4.7	sim
lt0m1m3	m1m3 control	LCU-68k	15%	6.2	sim
lt1adc	adapter rotator	LCU-68k	90%	6.7	real
		LCU-PPC	15%	6.7	real

Details of CPU load:

la0agc: 74% = 16% ccdcon, 14% ccdip, 1% ccdit, 15% ccdrdt, 9% interrupt, 5% agGuiding, 4% tNetTask, 2% lccTime, 2% tLqs
lt0alt: 50% = 16% servo task, 10%, fgTask, 2% bgTask, 7% Interrupt, tintTIM, 1% lccTime, 1% lccLogger
lstrap: 48% = 18% rdbServer, 2% strapServer, 2%, lccTime, 2%lccEvent, 4% Interrupt, 2% tLqs
(measures taken on LCU-68k-13MB, when strap was integrated)

Estimations for ATCS

The following table gives (conservative) estimates of the CPU loads and memory usage of the ATCS machines. All LCUs will be equipped with a Motorola MV2604 PowerPC 333MHz CPU board:

Machine	Description	Type	CPU-load	Memory Usage
wx<n>ats	Telescope Control	WS-J2240	65%	N/A
lx<n>alt	Altitude LCU	LCU-68k	70%	8MB
lx<n>az	Azimuth LCU	LCU-68k	75%	9MB
lx<n>aux	Auxiliary LCU	LCU-68k	30%	6MB
lx<n>dcs	Sensor LCU	LCU-68k	164%	24MB
		LCU-PPC	60%	24MB

For the Sensor LCU two alternatives are shown: obviously a 68k is not suitable.

lx<n>dcs(68k): 74%(ccd, ag)+40%(FAS/TSD tracking)+10%(chopping)+40%(strap) = 164%
lx<n>dcs (PPC): 60%

The only critical part is the Sensor LCU. Not so much concerning calculation power, but interrupt handling, database handling and messaging. More concerning is here, if everything can get correctly synchronized and tasks are not blocked due to some heavy sharing of LCC tasks.

Last modified: Fri Jul 7 14:30:04 METDST 2000