Minutes of the LArg Front-end Optical Links Meeting

Present :

Marie-Laure Andrieux    ISN
Alexandre Chekhtman	Novosibirsk
Jacques Colas		LAPP
Bernard Dinkespiler	CPPM
Olivier Martin		CPPM
Robert McLaren          CERN
John Parsons		Nevis
Mark Pearce             KTH
Erik Van der Bij        CERN (part-time)

SUMMARY
~~~~~~~

(1) Status of the ANALOGUE link project
	CPPM - O.Martin	
		Olivier reported that preliminary tests of the
		`n-up' Sandia VCSEL arrays indicated good stability
		and efficiency. Two of these VCSEL's would be placed
		into the ISN `demonstrator' packaging (see below). 
		Results from these devices would form the basis of
		a final paper on the analogue optical link development.

		Olivier also remarked that he had been in contact with 
		K.Lear at Micro-optical regarding commercial costs for
		VCSEL arrays developed by Sandia. If a new mask had
		to be built for the long-lead (approx 1 mm.) versions 
		that we use, the prices (for a quantity of 2000) would 
		be $100's per array ! The short-lead (approx 100 um) 
		versions, however, are in stock but no price was quoted. 
		The question of whether	we can arrange for the Sandia mask 
		to find its way to Micro-optical was raised - as we have,
		in principle, paid for it's development. Olivier agreed to 
		look into this.

	ISN - M-L.Andrieux
		Marie-Laure reported that the two `best-candidate' VCSEL 
		arrays would be incorporated into ISN demonstrators. Both 
		CPPM and KTH would receive one for final measurements.
		
		The status of the Amphenol connector was also reported on.
		Marie-Laure said that the mechanical tests on the ceramic had
		been completed and that electrical tests would be completed 
		in January. By the end of January, optical and, in particular,
		alignment tests would be started. Amphenol plans to have a 
		complete link prototype ready by the start of March. 	

	KTH - M.Pearce		
		Mark reported on some measurements performed with Jesper
		Söderqvist on a Sandia VCSEL array passively coupled to a 
		MT-connector. This followed the procedure successfully
		used for the PIN-array in the receiver box. There were
		some problems bonding the VCSEL array as the metalisation
		came off the chip - only 6 from 16 channels were successfully
		bonded.

		The preliminary measurements with this passively aligned
		VCSEL were encouraging. Final results on dynamic range
		etc. will have to wait for the final ISN demonstrator
		but Jesper noted that at present the link's dynamic range
		was limited by noise (1mW/2uW=500 or <9 bits). Since there
		is a cover-glass over the VCSEL, this noise maybe reflection
		induced - this needs to be investigated. In summary, 	
		passive alignment was shown to be possible even though
		the ceramic used was optimised for PIN-arrays which have
		a larger active area.

		Finally, Mark said that the KTH receiver box would soon be 
		tested	with fast transimpedence amplifiers at GHz speeds,
		if possible.

(1) Status of the DIGITAL link project
	Preliminary Results from October SARA Irradiations - M.Pearce
		Mark presented preliminary results prepared with
		Johan Lundquist on the neutron irradiation of commercial
		VCSEL's from Honeywell (9 off) and Mitel (9 off). Both were 
		housed in ST-style capsules. Through-out the irradiation, the 
		VCSEL's were biased above threshold. During the irradiation, 
		P-I curves were taken for each VCSEL at regular intervals.
		Mark presented plots showing that the VCSEL threshold currents
		did not move significantly during the irradiation. He also
		showed that, on average, the VCSEL light output at a given
		current dropped by approximately 10%. The VCSEL's were biased
		above threshold after the neutron beam was stopped. Annealing
		behaviour was observed and the pre-irradiation light output
		was recovered. The VCSEL's received an estimated total fluence
		of approximately 1.7x10^13 n(1MeV)/cm^2.

		Mark also reported on the irradiation of Hewlett-Packard ECL 
		G-link serialisers prepared by Bernard Dinkespiler. During 
		irradiation, only the PLL (analogue) part of the chip was 
		exercised. No loss of PLL functionality was observed during 
		the test. Back in the lab, Bernard tested the integrity of the
		chips's data-way up to speeds of 200 MHz - no problems were 
		observed and the PLL worked with output frequencies up to 
		1.6GHz. The chips received an estimated total fluence of 
		approximately 0.8x10^13 n(1MeV)/cm^2. It should be noted that 
		since there was no data-flow through the chips, it was not 
		possible to check for SEU or SEL etc.	
			
		Finally, some results from the irradiation of commercial 8-way
		multimode fibre-ribbon were shown. The ribbon was supplied
		by ACOME and a 10m long section was irradiated. Initial 
		results indicated an attenuation of between 20-30% due to the 
		irradiation. There was some recovery after the irradiation was
		finished, and the exact nature of this was still being studied.
		The fibres received an estimated total fluence of 
		approximately 0.5x10^13 n(1MeV)/cm^2.
 
		In all cases, more sophisticated tests are planned in the
		future. For example, with Gb data flows through components 
		and BER tests etc.

	News from Co-ordination Meetings - M.Pearce
		Mark reported on the advent of an overall ATLAS co-ordination
		working group for front-end links. Full details can be found 
		from : http://www.cern.ch/Atlas/GROUPS/FRONTEND/links/

		In particular, Mark pointed out that the working group has
		requested workplans to help assess commonality between 
		subdetectors and that the electronics co-ordinators had
		requested some uniformity for link costs within CORE.

	Timescale and Milestones - M.Pearce
		In the spirit of the ATLAS links working group, Mark presented
		a preliminary draft of a workplan and associated milestones
		for LArg links in the coming 18 months. Mark expressed a desire
		to develop this into a more formal document in the New Year,
		to be presented to the LArg community as a whole. To set the 
		scene, he indicated the following milestones :	
			o Freezing of front-end system (data-flow) architecture
			  (see talk by John Parsons)
			o Choice of link architecture. Parallel or serial.
			  EMC considerations between preamps and serialiser
			  clocks.
			o Definition of a neutron, gamma, other (?) irradiation
			  programme
			o Definition of a demonstrator link programme
		He also discussed the need for some planning of accelerated
		aging tests and interface to installation engineers (see talk
		by Mark later).
 
		In order to meet these milestones, Mark thought that given
		the geographical separation of the groups involved it would
		be sensible to define distinct projects.
		A list of possible project areas was presented based around the
		need to be able to perform realistic function tests during 
		irradiation and aging tests, for example. These projects would
		be refined in time for the next meeting.

	Proposed Link Baselines - B.Dinkespiler	 
		Bernard presented plans for Gb demonstrator links and test-beds.
		A serial version used HP G-link Gb serialisers with 
		inter-changeable optics, such as : ST-housed VCSEL's + drivers
		or commercial transceiver modules. The PCB's for this project 
		were expected to be received back from industry very soon. 
		Bernard also presented an idea for a pseudo-random BERT based 
		on a pattern generator and a FIFO-buffered Altera-based error
		detector (see the transparencies for a clearer picture !).

		A parallel version using, in the first instance,
		an Optobus as the optical elements is also being
		designed. In this version, there are 4 data lines and
		one clock line at 320 MHz to strobe the data out.		

		Bernard said that these set-ups could be adopted for EMC (using
		the test-beam electronics with no beam) and irradiation tests.

	Radiation Testing Issues - M-L.Andrieux
		Marie-Laure presented details of the SARA time schedule for
		1998. Since SARA closes at the end of July 1998 it is 
		essential to fully use this facility next year. A preliminary
		testing schedule was shown and will be refined over the
		coming months. She also explained that post-SARA neutron
		irradiations may be possible at Orleans.

	A Gamma Irradiation Facility in Stockholm - M.Pearce
		Mark reported that Johan and he had visited a hospital in
		Stockholm which had a Co-60 gamma source. Dose rates of 2.1 Gy 
		per minute were possible @ 60cm from the source. The beam area
		at this distance is approximately 20cm x 20cm. The source
		was only accessible at the weekend and during a typical 70 hour
		period, a total dose of 8820 Gy was possible, which is 
		perfect for LArg testing needs. Mark and Johan would carry 
		out tests, like those reported on earlier in the meeting
		at SARA, in January to gain experience with this new facility.

	Comments on LAr Link Architecture - J.Parsons
		John reported that the baseline architecture for detector
		control and TTC distribution has :
			o one control card per half front-end crate for
			  fanning-out and distributing clock (point to point), 
			  TTC and control (bussed) to each front-end board 
			  (FEB). 
			o a single link for bringing TTC information to each
			  control card (from TTC system)
			o a single link bring control words (from the ROD)
		John pointed out that this scheme has many single point
		failure modes which could lead to the loss of half a
		front-end crate (1792 calorimeter channels). He noted that it 
		would perhaps be better to bring separate links per FEB, ie: 
		one link for TTC and another for serial control. This could be 
		done optically, or with copper as the data speeds are rather
		modest (approximately 5 Mb/s for control and 200 Mb/s for TTC).

		For FEB-ROD links, there are 128 calorimeter channels per FEB
		each running at a nominal 1.28 Gb/s. Each FEB serves a single
		depth segment (ie: PS, strips, middle or back). However, the 
		L2/DAQ requires that trigger towers (SUMMED over PS, strips, 
		middle back) are available in a single ROD. A remapping of 
		calorimeter channels onto trigger towers between FEB and ROD 
		could be achieved by re-grouping into sets of 32 channels, 
		rather than 128. In fact, the FEB architecture is designed such
		that a 32 bit word is composed of 16 two bit contributions from
		separate ADC's (each ADC serves 8 calorimeter channels). So, 
		one could naturally achieve the re-mapping using 4 links per 
		FEB (8 for the presampler). This also means that only 32 
		calorimeter channels would be lost if a single link went down. 
		Of course, the cost of such links is a major concern when 
		considering such a strategy, given the extra number of links.

		After this presentation there was some discussion. In particular
		it was noted that the TTC/control and FEB-ROD links could
		effectively be decoupled as the responsibilities lie with
		different people, ie: the the TTC/control links are the 
		responsibility of the trigger/DAQ group. This means that the 
		FEB-ROD links remain unidirectional with only the
		emitters in a high radiation environment. 

	Installation Issues - M.Pearce
		Mark reported on a meeting with Mark Hatch regarding the	
		installation of LArg FE-link optical fibres. Mark H. said
		that the installation team would like to have fibre routing	
		drawings completed for the installation TDR in September 1998.
		Several points, in particular, were discussed :
			o In the TDR it is assummed that a single fibre will
			  be run from FEB to ROD. Due to the high radiation
			  environment in the crack region, this may not be
			  possible. If a special radiation hard fibre were to	
			  be used in the crack then a patch panel would be
			  needed somewhere to allow connection to a more 
			  standard `off the shelf' (cheaper !) fibre. The
			  position of this patch panel needs to be accounted 
			  for in an already congested part of the detector.
			o Mark H. had suggested that if many FEB's were routed
			  into individual fibres housed in a ribbon then it
			  may be sensible to have the option of replacing
			  fibres in case they are broken during access periods.
			o In order to allow designers to proceed with the 
			  allocation of space for fibres, a worse case scenario
			  was presented. Assuming there are 30 FEB's per crate,
			  each with a maximum of 4 data connections and 2 
		          TTC/control connections, implies a total of 180 
			  fibres per front-end crate. These fibres could be 
			  supplied in ruggedised ribbon-cables.  
		There was some discussion after this presentation : Where a 
		patch panel for fibre type conversion is needed, then there is
		the possibility of doing the trigger tower `remapping' here.
		The additional size needed for the patch panel would need to 
		be evaluated though. Jacques asked if in the case where the
		ROD is placed in the control room, 200m from the detector, 
		whether continuous fibres could be over the full distance
		or whether an extra patch is needed. He also asked what the
		minimum length a fibre can be when fanning out signals from a 
		central point to avoid reflection problems and ensure mode
		mixing.

TRANSPARENCIES
~~~~~~~~~~~~~~
	Copies of transparencies will be sent to :
		o A representative from each of the participating
		  groups (CPPM, ISN, KTH, SMU)
		o The ATLAS secretariat