Summary of Meeting at Oak Ridge National Laboratory on Muon Tracker Electronics on May 16, 1996

Compiled by D.M. Lee and W. Sondheim (phenix-muon-96-6; submitted June 3, 1996)

Attendees: Ken Read, Glenn Young, John Halliwell, Ed Kennedy, Jim Walker, Alan Wintenberg, Walt Sondheim, and Dave Lee.

A meeting was held at Oak Ridge National lab on Thursday, May 16, 1996 to discuss the integration of the muon tracker electronics with the mechanical structures at station 2. Present were Ken Read, Glenn Young, John Halliwell, Ed Kennedy, Jim Walker, Alan Wintenberg, Walt Sondheim, and Dave Lee. A number of issues were discussed concerning the space restrictions imposed at station 2, possible mounting structures at station 2, present state of the electronic developments with ADC's and AMU's and the impact of the CSC electronics, and limitations imposed by the present understanding of the DSP/DCM's.

Mechanical Constraints and Electronics Support Structures

Dave Lee brought drawings of the present station 2 support structures with details of the space available at the outer circumference where the support structure is near the lampshade. It was decided that we would extend the short tie bars at 45 degrees that tie the front support structure with the back support structure in 'z' to provide a means to attach the electronics. Cross bars between the tie bars will be used as the electronic mounts. Since these ties bars and the support structures are all held at the same potential as the octant chambers, i.e. ground and isolated from the magnet, electrically this would be good for the electronics. The space between the 35 degree line and the lampshade is 16.5 cm. Allowing for a 2 cm stay clear from the lampshade all of the electronics must reside in 14.5 cm. The guidelines for this were that the front support structure and the tie bars are permanently installed and for installation and removal of chambers the back support structure is removed and/or reinstalled. LANL will incorporate this in the mechanical designs. A brief description of the approximate location for the seven lines of sight for the optical alignment monitors was given, three locations along the top and four along the sides.

PROPOSED ACTION: LANL will incorporate the lengthened tie bars and electronics in its mechanical drawings.

Mother Boards,Daughter Boards, System Requirements, etc.

A general discussion of electronic issues outlined some of the problems, and limitations that are important to understand in the full integration of the electronics and mechanics on day 1 and later. Much of the following came from Glenn, Ken, and Alan.

10% working spares of all motherboards and/or daughter boards is mandated.
The DSP input from the G-Link will run at 40 Mhz or 20 Mhz. Must determine which is correct.
The G-link is a 21 line ribbon cable that can operate in a 16-17 bit mode or 20-21 bit mode at up to 1.5 Ghz. Expect to operate at 900 Mbits/sec at cost of $500. We will most likely use 16 bit mode.
FEM must be scalable by 2 to include the possible increase in the number of DCM's. (future upgrade)
ADC speed is 100 Mhz or 144 Mhz with clocking on both transitions. The conversion times and the number of samples in 40 us are,

clock	#bits	time	samples
100Mhz	11	10.2us	3
144Mhz	11	7.1us	5
100Mhz	12	20.4us	1
144Mhz	12	14.2us	2

An 11 bit ADC exists at ORNL but not a 12 bit. A 12 bit ADC requires development. The anticipated ADC/AMU combination will be 32 channels. The possibility of adding two (or more) ADC's per channel also requires development because the AMU can't handle multiple ADC's.

Dave Lee says we need a minimum of 11 bits. For the rare event where amplitude is very large and saturates the ADC, next nearest channels can be used instead.
According to Leo P. on day 1 the Muon Tracker will have 70 DCM modules that will handle 280 fiber links. That means about 140 fiber links per muon arm. 98% of these links will be used for the fine resolution cathodes. According to Leo P.'s chart from the ONLINE meeting at Oak Ridge, we will have the final number of DCMs on day one.
The maximum # of words the DCM's can take is 1600 for 40 Mhz transmission and 800 for 20 Mhz transmission.

The number of samples possible depends on the transmission rate, the number of channels per link, and the ADC clock speed. Ed Kennedy is mostly worried about noise levels and would like to see as many samples as possible. We all agreed that the cable between the cathode strips and the front end amplifier must be a shielded cable. Since an 11 bit ADC/AMU is available and acceptable as the minimum needed we decided to propose that we would settle on an 11 bit ADC/AMU with 5 samples. To accomplish this for a 40 Mhz transmission rate we need,

(16000 channels/140 links ) x 5 samples x (1/40 Mhz) = 14.3 us

A transmission rate of 20 Mhz would be adequate.

PROPOSED ACTION: 11 bit ADC/AMU with 5 samples

The grouping of fine cathodes on each mother board was discussed at length. The "no phi grouping" original request creates serious problems in board layout and cost so we agreed to propose of dispensing with that requirement and allow any grouping of fine cathodes that would be cost effective. The grouping that was most appealing was the three cathodes would be read out on the same motherboard. The number of channels per motherboard would be decided by cost and transmission rate. It was felt that board dimensions would be in units of sixteen channels and the most logical segmentation would be 96, 144 or 196 ( the count used in the WBS). For the 144 channel solution and 5 samples this corresponds to 720 words and for 196 this corresponds to 960 words. The 144 channel solution would work with 20 Mhz transmission rate, the 196 would not.

PROPOSED ACTION: No limitations on the grouping of fine cathodes.

The motherboard /daughterboard concept was discussed. Walt and Dave questioned and Ken, Alan, and Jim emphasized modularity, testing, and replacement. The results centered on the difference between connectors and cables as opposed to board to board connectors. Board to board connectors similar to the connectors on VME crates were felt by ORNL to be very reliable and not prone to the same problems that exists with ribbon cables and connectors. In addition the mother/daughter combination is appealing for its modularity where each individual major component can be tested separately and replaced if necessary without resorting to replacement of the whole mother board. The arguments for the mother/daughter concept were persuasive and the use of board to board connectors with no cables was believed to be sound. Cost was not thought to be an issue. The daughter boards could be simply the front end preamps and ADC/AMU. The links and heap manager could be other daughter boards.

PROPOSED ACTION: No cable interconnects between mother and daughter boards, board to board connectors are O.K. Mother/daughter concept preferred.

The anode readout (and low resolution cathode readout) is only 1 bit per channel and is widely spaced. Having a bus to readout all the anodes will be limited by the maximum buss speed of about 10 Mhz. To preserve the pipe line nature of the system a 16 channel FEE card would form a sixteen bit word with each bit corresponding to the individual channel latch and the output would be a ribbon cable routed to a central motherboard at the outer circumference of the octant. The amplifier cards would be along each spoke. All anodes in an octant would be readout by one motherboard. For station 3 this would amount to 66 words of 16 bit data per octant (also 66 ribbon cables of varying length from bottom to top!!). More than one octant could be readout by one mother board to more effectively utilize the fiber link if the routing comes to a common point for adjacent octants. No action was taken on the anode readout other than to establish that the anode FEE would be along the spoke. More thought is required.

The low resolution cathodes are also sparse and could be readout by the fine cathode mother board. The overhead in the number of words added to the fine cathode data stream is very small. A separate low resolution daughter board would be used but otherwise the motherboard would not be changed. Since there are fast clocks already on the motherboard and fine cathode daughter boards it was felt that the addition of a few latches would not effect the fine cathode noise. We all agreed that this would be a good approach.

PROPOSED ACTION: Try to incorporate the low resolution cathodes onto the high resolution motherboard.