CSC and Overlap Muon Track Finders

Re-Design During Project Review

There was a great deal of thought given to the design of the CSC and CSC/DT overlap Muon Track Finders during the late-97 /early-98 descoping of the US-CMS project. We lowered the base cost of the project by 25% (material cost by nearly 50%) using the following simplifications:

Muon Port Cards now receive signals from chambers covering 60 degrees in phi, whereas before there were alternating 20 and 40 degree MPCs. This lowers the number of MPCs, optical links, Sector Receiver cards, and Track Finder crate backplane signals and cables. The 60-degree boundaries are matched up as well as possible with the barrel: this happens at 15, 75, ... degrees. It is suggested that the barrel stub information should also be collected in 60-degree segmentation to allow similar cost savings and maximum compatibility with the endcap. (There is one exception - the ME1 data is collected in 30 degree sectors and combined into 60 degree sectors by the Sector Receiver cards, due to the high number of chambers in this station.)
A critical data flow bottleneck in the CSC muon Track Finder system has been reduced by nearly a factor of three by ignoring those muons which cross sector boundaries. This introduces cracks in the muon trigger acceptance which are very small, because of the wider sectors and the small bending of muons in the endcap region. Also, the cracks narrow inversely to the PT of the muon, i.e. they are smallest for the most interesting class of muons.
Only three CSC stations are assumed, according to the de-scoping of the endcap muon system. This further reduces the number of Port Cards and optical fibers. However, in order to allow eventual re-scoping of the system, provision is made in the Track Finder to handle four stations.
The CSC Track Finder now fits into 8 VME 9U crates (previously 24 crates). As shown below, these crates are arranged in four racks: for each muon endcap, there is one rack for CSC-only track finding and one rack for CSC-barrel overlap track finding. Each crate handles muon track finding within a 180-degree swatch in phi. Despite the contraction, the design allows room within the crates for expansion of the Muon Track Finder to handle sophisticated tracking algorithms.
Handling three stubs per crossing per sector rather than only one stub per sector is not so expensive, of order $100K. If studies indicate that 4 stubs/xing are required, this may be possible. This is clearly an important area for muon trigger simulation work, but we also need to coordinate design with the barrel (where, e.g., only two stubs per 30-degree sector are now contemplated).

The following sections present details of the muon Track Finder layout and data flow, with special emphasis on the CSC portions.

CSC and Overlap Muon Track Finders: Block Diagram

Here is a CSC-centric block diagram showing the electronics between the stubs sent by Trigger Motherboards (TMB) to the Muon Port Cards (MPC) and the tracks sent by CSC and Overlap Sector Processors to the global muon trigger:

There is a .pdf version available here, and the original Persuasion file here in Macbinary format.

VME crate layout of the muon Track Finder

The layout of the muon Track Finder contains separate racks for barrel (DT), endcap-only (CSC), and overlap processor regions. Within the barrel system, the Track Finders are organized by wheel. Each VME crate handles 180 degrees in phi of track finding. Signals from the CSC system come to the Track Finder on optical links, and are received in the CSC-only section. A conception of the system, shown below, contains 9 racks, each of which contains 2 VME crates. The CSC-only and overlap Track Finders (U.S. responsibilities) are contained in 4 racks/8 crates.

Here is the same picture, but scaled by the rapidity coverage:

Inter-crate Data Flow

Connections between sectors in phi need to be made in order to accomodate bending tracks which may travel from one sector to another as they penetrate the several muon stations. This is done in the Barrel DT system, but not in the Endcap CSC system. The justification is that the bending is small in the endcap region for a fixed Pt. Also, the cracks which are created are smallest for the most interesting muons, i.e. having the highest Pt.

There need to be "rapidity" connections between the various muon Track Finder crates because the boundaries are non-projective. In the Barrel DT crates, keyed off of the rapidity of MB4, the outer layer, information from MB1,2,3 flows toward the ends: wheel 0 sends to wheel +-1, and wheels +-1 send to wheels +-2. Wheels +-2 send information to the overlap crates. The CSC-only crates, which receive the information from all CSC stations, send information to the overlap crates. This is shown below:

CSC Sector Receiver Cards

Data comes from the Muon Port Cards (MPC) on optical fibers to the CSC-only Track Finder crates. Each MPC sends data representing 60 degrees in phi of one muon station (with the exception of station 1, which sends data in 30-degree sub-sectors). Each muon stub which is transmitted requires 36 (+- a few) data bits. The card which receives these signals is called the Sector Receiver (SR). It is assumed that HP Glink chip set can transmit 21 bits per 25ns bunch interval, of which 5 are used for error detection, thus leaving 16 bits of useful information per BX. The data bits per muon stub are counted as follows:

Cathodes (18 bits)

3 bits - Front End board index (0-4)
5 bits - 1/2-strip ID
1 bits - L/R bend
8 bits - Pattern number
1 bits - High/low Pt flag

Anodes (14 bits)

3 bits - Front End board index (0-6)
4 bits - Wire gang
7 bits - Wire pattern

Other (4 bits)

4 bits - Chamber index (0-8)

Of course, there are a few other signals such as clock, BX0, which are transmitted, but most of these are common and will be stripped off by the Sector Receiver cards.

As an (simplistic) example, suppose only one muon stub per crossing is sent by each MPC. Then two optical fibers per station carry this data to a Sector Receiver which receives either 8 or 10 fibers (depending on whether we end up with 3 or 4 muon stations), and 144 or 180 bits per crossing.

The Sector Receiver card does 2D to 3D muon stub conversion, alignment corrections, and data reformatting. It then sends muon stub data to the Sector Processor (SP) module on the crate backplane. If the Sector Receiver is being used in the CSC-only crate, it must also replicate the data and send it on to the Overlap crate (probably via copper). If the Sector Receiver is being used in the Overlap crate, it must receive its input data over copper. We therefore design a single card having both types of functionality according to the following data flow diagram:

After the data is received by the Sector Receiver cards, it can be reformatted by look-up tables into whatever format is convenient for the Sector Processor track finding. This will generally result in somewhat fewer bits transmitted per muon stub than the 36 bits which are received on the optical fibers.

The phi position is given by the 18 cathode bits listed above plus the 4 bits which indicate which chamber the muon passed through. However, the minimum strip width in the CSC system is 0.125 degrees, and The limit of trigger resolution per muon stub is about 0.1 strip widths, or 0.0125 degrees RMS. Using 12 bits for phi position within a 60 degree sector, the resulting bin size is 0.0146 degrees, and the RMS error is 0.0042 degrees. This is 3 times smaller than the the intrinsic resolution, and adds only 5% to the phi resolution when added in quadrature.

Using 11 bits to indicate rapidity allows for precision of (2.4-0.85)/2048 = 0.00076 units of rapidity, or RMS of 0.00022 units of rapidity. This compares to the maximum precision of about 0.00054 units of rapidity from the wire gangs. The rapidity binning is 2.4 times smaller, and added in quadrature, adds only 8% to the rapidity resolution.

Finally, the phi direction of the muon stub (local "bend angle") may also prove useful for improving the momentum resolution of the muon Track Finder. We can use one bit to indicate the sign of the bending, plus an additional 5 bits to indicate the magnitude. The 5-bit magnitude can be 0-15 for stubs with bending from infinite momentum down to the nominal high-Pt 10 GeV limit, and 16-31 for stubs with bending from 10 GeV down to 2.5 GeV where the low momentum muons range out in steel.

As in the Barrel DT case, there are an additional 3 quality bits whose exact definition is to be determined later.

By the above counting, the maximum number of bits which need to be sent from Sector Receivers to Sector Processors is 32 in the CSC system - 12 phi position bits, 11 rapidity bits, 6 local bend angle bits, and 3 quality bits.

Data Flow Within the CSC Track Finder Crates

There are three Sector Processors per crate, each handling 60 degrees in phi. In the baseline design, up to three muon stubs are transmitted from each MPC on six optical fibers. The Sector Receivers are designed to receive 15 optical fibers carrying 270 data bits. Depending on whether we have three or four endcap muon stations, two or three MPCs are connected to each Sector Receiver. The Sector Receiver, which is connected to the two ME1 MPCs in the 60 degree sector, combines the information before transmission to the Sector Processor. Two Sector Receivers send data to each Sector Processor (288 or 384 bits, depending on whether 3 or 4 muon stations are built). One also needs one ground connection per approximately eight signal connections to the backplane. One therefore requires high-density Z-pack (Futurebus+) backplane connectors- these pack 192 pins in 96 mm vertically. (About 350mm are available vertically on a 9U (366mm) height Eurocard.) The worst bottleneck, the SP backplane pin count, with 4-station operation, is shown:

3-stub, ME1-4

In this case, three high-density connectors are required, using up most of the backplane connector space. A non-standard readout bus will need to be defined.

Data Flow Within the Overlap Track Finder Crates

In the 4-station muon system, data from muon stations ME1-3 and MB1-3 are sent to the Overlap Track Finder crates. This requires 96 bits (x number of stubs/MPC) from each sector of the CSC system, and 66 bits (x number of stubs/station) from each sector of the Barrel DT system. In any case, this requires use of the high-density Z-pack (Futurebus+) connectors. One needs a special module to receive the barrel signals, which is built by European groups. In the baseline design, up to three muon stubs are transmitted from each MPC on six optical fibers. In this case, two SR modules receive the CSC data as explained above, plus another "SRDT" module (or possibly two) receives the DT signals. This crate starts to get rather full as shown below. Again, the high-density Z-pack (Futurebus+) connectors are used to connect to the backplane, and nearly all of the backplane space is used, and a non-standard readout bus will need to be defined.

3-stub, overlap

N.B. We have not considered clock, control, and readout signals. In order to collect these signals as well as more data signals (if needed) at the Sector Processors, there are three potential ways to do it:

Higher-density connectors (what?)
Use of the front panel. This makes access to the cards difficult, whether required for replacing modules or debugging them in situ.
Add a mezzanine board on the Sector Processors to access more signals from the backplane. This uses up several more VME slots.

Architects of the CSC Track Finder

And many thanks to Wesley Smith, Grzegorz Wrochna, Alex Kluge, Torsten Wildschek, and JK Smith.

Return to CSC Trigger web page.

This page is maintained by

Jay Hauser (hauser@physics.ucla.edu), last updated 27 January 1998