|CERN -> ATLAS -> TU Vienna|
|High-Density Digital Links|
Defining the potential of the conductive structures and building a signal-reference system inside the TRT, can be chosen within two philosophies: either strongly connecting everything together or trying to control the currents which flow in the systems. The first one yields the lowest impedance between any two points of the system, but simultaneously allows loops and shield currents to flow inside the system. The second approach allows to break this loops and to ban shield currents from intruding the system through carefully provided low-impedance paths.
In addition each sub detector has to follow "The ATLAS Policy on Grounding and Power Distribution"  which gives the following guidelines:
Figure 3.22 shows the grounding scheme for the TRT end cap. The smallest unit of the end cap is 1/32 of one half of a wheel which is created by the granularity of the front-end boards and their power supplies. The ASDBLR and DTMROC boards have multiple signal connections between each other. Thus their ground planes are full planes which are connected tightly over multiple ground pins on their interface connectors. The power for these boards derives from a common input connector at the DTMROC board. The modularity for the power supplies is 1/32 of a complete wheel. Thus, the two halves have to be decoupled by the impedance Z of the power-supply filter.
The front-end electronics plugs onto the active webs. The cooling structure connects the two active webs and the two passive webs and provides a substantial ground plane. A copper-plated kapton sheet connects the inner seal and ring 1, which are capacitively coupled to the wires of the straws, to the ground plane of the active web. Because this connection is seen as a part of the signal return path, the "left" and "right" side are connected and force the return current of the "right" side to flow over the left "side". This is undesirable, whereas breaking this connection and adding a return path for both sides would be a more uniform approach. Ring 2 and 3, which are strongly coupled to the outputs of the straws, are also connected in this way.
A segmentation of the system into ground subsystems in the azimuthal direction is impossible because of the unitary carbon fibre rings and the cooling and gas pipes. In this case the different parts of the ground system have to be connected together to provide the smallest impedance.
In the axial direction a segmentation, which makes it possible to control the system currents, is possible. Though there is a high capacitive coupling among the wheels - especially between two copper-plated kapton sheets between two wheels - a separation can decrease crosstalk. Nevertheless, a common reference point is necessary to stay in the common mode range of the differential receivers for data and control signals. The squirrel cage would provide a good conductive structure for this purpose. As the squirrel cage is intended to be part of a shielding structure, shield currents may introduce potential differences. In addition, the squirrel cage is not present at the level of the wheels C. The reference point should be chosen to provide the same potential for the front-end electronics and the structures which surround the sensitive part of the detector. In the case of the TRT this would be the ASDBLR-board ground. The ground planes of the connected ASDBLR boards would build the reference ground plane for the system. For feasibility reasons, the level of the web and the cooling structure was chosen as reference plane. Additional ground connections between an ASDBLR board and the active web strongly bind the ASDBLR board-ground to this reference. Connecting the active webs to its neighbors of the next wheel with a big number of small value resistors - three per 1/96 of a wheel - and additional ones in the azimuthal direction among the flexible parts of the webs provide a good global reference plane and allow local filtering of voltage differences .
The wheels connected in this way to one module with a common shield have only one single connection to this surrounding shield. The safety and reference-ground cable connects at the same point from the outside. This provides the reference and safety potential and avoids shield currents flowing into the system. An impedance can decouple the reference plane for high frequency as no current should flow over this connection.
The design of the grounding scheme has also to include the possibility that one could change to the second grounding approach, namely solid ground connections. This requires the replacement of the resistors by direct connections, the possibility to connect the inner seals together, and additional connections between the DTMROC boards and the squirrel cage or over-all shield respectively.
Assuming that the ATLAS grounding scheme could be changed, we have to plan additional possibilities to connect the TRT to the cryostat.
Figure 3.23 shows the grounding scheme for the TRT barrel. The straws of each module are housed in separate carbon-fibre shells. The front-end electronics covers both ends of the shells. A copper layer at the outer wall of the outermost module and one at the inner wall of the innermost module build, together with a special shield layer at the interconnected roof boards, an over-all shield.
The roof boards will also build the reference plane, to which the reference/safety line connects. All of the electronics inside the shield tie tightly to the reference plane. The ground plane of the tension plate connects through the cooling plate and the ASDBLR and DTMROC boards to the reference at the roof board.
The incoming services are decoupled over the filter impedances.
This grounding scheme foresees to have the lowest possible impedance between any two points of the system. This will allow currents to flow over multiple parallel paths. Even shield currents can enter the system and change the local ground reference.
However, the current grounding scheme will have to change with the potential new design changes mentioned in Chapter 3.3.4. E.g. a flexible roof board will not provide the demanded shielding functionality. Adding a copper-plated kapton sheet to close the over-all shield instead of the using the roof boards would decouple the reference function from the shield function. Introducing a cable-tray concept like for the end cap would allow one to reduce the number of shield penetrations and add an additional noise filter.
|data & control signals|
Figure 3.24 shows how to connect the data and the control cables to the grounding system. The individually-shielded twisted pair (STP) is connected directly to the DTMROC (or roof respectively) ground and over a capacitor to the local ground of the active patch panel at PPB/F2 to break the ground loop for low frequencies but to allow to short-circuit high-frequency noise currents. The common shield of the twisted-pair bundles (UTP) is connected directly to the patch-panel ground and over a capacitor to the ground of the back-end electronics. Separate cables or braids respectively provide the reference point and safety.
The cable tray connected to the over-all shield acts as feed-through capacitor and second shield for the cables. It is not directly connected to the individual shields of the cables. In addition it cancels the loop area of potential loops between two cables. Additional cable trays after PPB/F1 applied in the same way are recommend but not yet considered.
The low-voltage-power-supply cables are not fully defined yet. Preferably, the analogue and the digital supplies should have separate ground returns and individual shields - see Figure 3.25.
The shields are connected directly to the DTMROC ground plane and over a capacitor to the low-voltage patch panel at PPB/F2. Thus, it will not allow low-frequency currents to flow. High-frequency currents are short circuited over the capacitance to the cable tray. The lower static magnetic field and the higher available space than within the Inner Detector allows one to use efficient filters at PPB/F2. Thus it is possible to switch to cheaper multi-core cables with a common shield. However, it is recommended to separate the analogue and digital supply lines. The common shield of these lines has to be connected to the safety line at the floating power supplies.
Miniature coaxial cables are used inside the Inner Detector for the high-voltage distribution. The shields of these cables carry the return currents. Additional pins of the connector patch panel at PPB/F1 and PPB/F2 provide increased conductivity for these returns. Outside the Inner Detector, the transmission is implemented by multi-core coaxial cables. Separate lines are reserved for the current return. Thus the shield does not have to carry any return current.
|gas & cooling pipes|
The separation of the gas and cooling pipes inside the TRT is not feasible. Thus they were chosen to build together with the cooling structure and the active and passive web the reference plane. Insulation pieces electrically break the pipes where they enter the TRT-detector volume. Inside the detector, the pipes are heavily connected to the cooling structures. Outside the detector they are bonded to the over-all shield.
Further insulation pieces at several detector interfaces can allow one to define the separate parts of the pipes as "electrical" parts of other sub systems. In these cases, the pipes are to be tied to the local ground.
A systematic approach will indicate where the connections belong in order to allow return currents to flow back to the source without detour.
An impedance approach requires one to provide the smallest possible resistance and reactance in the system. This means that one must put multiple connections in parallel to decrease the impedance.
In terms of frequency, it is a rule of thumb to make the connections each 1/20 to 1/10 of the wavelength of the highest frequency produced or introduced into the system in order to minimize standing waves.
According to this, the connections between parts in the TRT should be placed every 5 to 7 cm. Interconnection wires should be as short and as broad as possible for lowest possible inductance.
|February 9, 2000 - Martin Mandl||Copyright © CERN 2000|