Controls Interface to Electronics

Controls Interface to Electronics

Controls Interface to Electronics Boards Credit-Card PC and Local Control bus Beat Jost Cern EP Outline Introduction Goal Current Ideas for Solution Discussion and Conclusion FE/Controls/DAQ Workshop Beat Jost, Cern 2 Problem Introduction Classical Way of controlling electronics in HEP Parallel Bus (VME, Fastbus,) Ethernet WS Crate Controller (CPU) Electronics Modules Pros: Universally available

simple? slave interface in the past bus could also be used for DAQ FE/Controls/DAQ Workshop Cons: expensive CPUs (very small market) expensive crates expensive slave I/Fs Beat Jost, Cern 3 Assumption for the Following The vast majority of the electronics boards are home-made Crate bus is not used for moving the physics data in LHC experiments, because (at least) at higher levels of the readout systems, performance is insufficient trigger rate too high for processor intervention on per event basis Crate bus is only used for control and monitoring for this purpose a high performance bus is not needed for this purpose a parallel bus is not desired for reliability reasons One participant on the bus can prevent bus accesses even if its not involved makes diagnostics more difficult Hence A more reliable and perhaps even more cost effective alternative is desired. By the way: Crates are useless for dispersed individual boards

FE/Controls/DAQ Workshop Beat Jost, Cern 4 Goals Get rid of parallel busses for controls Find a cheaper solution for per-board controls interface (slave) per crate intelligence, by taking it out of the crate formfactor Use commodity items Crates (no parallel bus needed anymore) Reduce crates to Mechanical support (Anti-Gravity device) Power Bus (could be arguable) Cooling (Fantray) Take advantage of large market (low price) Provide a common controls interface for ALL electronics boards in LHCb FE/Controls/DAQ Workshop Beat Jost, Cern 5 Current Ideas Replace Parallel Bus (VME, Fastbus,) Ethernet

WS Crate Controller (CPU) Electronics Modules By Or Network Switch/HUB even better Power Bus Power Bus Field-Bus e.g. Ethernet WS WS Electronics Modules FE/Controls/DAQ Workshop Beat Jost, Cern Field-Bus e.g. Ethernet Electronics Modules

6 Electronics Board Architecture The architecture of an electronics board could look like this: E.g. 9Ux400mm I/O FPGAs Configuration Monitoring Diagnostics Debugging ... WS LUTs Regs DSPs ADCs TDCs Etc... I/O Power Connectors

I2C (simple)Parallel Bus PCI Bus JTAG Controls Interface Field-bus Reset Note: No other interface!! Standard Application Specific FE/Controls/DAQ Workshop Beat Jost, Cern 7 Requirements for On-Board ECS Interface Sufficient bandwidth into each board (10-100 Mb/s) Cost per board must be low Uniform approach for all electronics (volume) for Ethernet the cost of a switch or hub port has to be taken into account 10 Mb/s hub port ~30 SFr. (today) 100 Mb/s switch port ~80 SFr. (today) Can be mixed!! Mechanical

minimal height (thickness) minimal surface Software support Low-level access libraries from WS to board components Tools supporting the programming of a CPU (if present) on the interface Reset of ECS interface without disturbing the operation of the rest of the electronics on board FE/Controls/DAQ Workshop Beat Jost, Cern 8 Possible Solutions under Study Credit-card PCs (e.g. smartModules) CPU Mem DRAM Bus Ethernet Periphery Complete PCs!! Typically based on Intel compatible microcontroller plus glue logic, could replace

partially PC/104 in the future. Keyboard, Comm, Screen, Disk, Floppy, LCD,... ISA, PCI, I2 C (Master),JTAG (Master) FireWire Firewire FW-XX Parallel Bus I2 C JTAG We also looked at PC/104, but it doesnt seem suitable mechanically FE/Controls/DAQ Workshop Beat Jost, Cern 9 Immediate Plans For SmartModules acquire one evaluation kit and one bare module Evaluation Kit with 486 acquired, porting currently Linux to it Final (586-based) module not yet available learn how to program the processor using the evaluation kit build an evaluation board with FPGA(s), memory, registers, LEDs, to study how to use the modules and whether they are suited for our application being designed now. Ready in ~3 months?

Should be able to give guidelines for designers by November FireWire try to understand the availability of interfaces/bridges for example: we know that there exists a Firewire to IDE chip. Can this be used? Similar (or same) evaluation board as above to prove usability FE/Controls/DAQ Workshop Beat Jost, Cern 10 SmartModule Evaluation Board Front-PanelLEDs for: Running NOT Running some bits driven from registers Ethernet Traffic 40 MHz Clk Latch Analog Input High precision ADC (12 Bits) Integrator/Shaper MIC Delay chip Data

FPGA (Altera ) Push Button (Run Acquisition) Memory (>2 MB) Parallel Bus 32 Address lines/ 32 Data lines (nonmultiplexed) Registers Interrupt FPGA (Lucent) (DMA) PCI Clk Inhibit I2 C VMEClk VME/ Glue Logic

SmartModule PLX 9080 Parallel Port JTAG (SW JTAG) FE/Controls/DAQ Workshop JTAG Glue Logic Beat Jost, Cern HW JTAG VME Slave Interface RJ45 Connector Ethernet Running 11 Local Bus Characteristics Based on PLX PCI 9080 chip (see also http://lhcb.cern.ch/computing/controls/pdf/9080db-106.pdf) Goal: It should be easy to interface to components to local bus }

32 bit address Non-multiplexed 32 bit data + Byte Parity Control lines (address strobe, R/W, Byte enable, etc.) Local bus clock frequency 0-40 MHz -> allows to run bus synchronous with LHC Clock local bus can be Little or Big Endian One interrupt line from local bus to PCI -> might need external interrupt register/generator not terribly fast: } 8.3 MB/sec PCI reading from local bus Single Cycle from timing diagrams 8.8 MB/sec PCI writing to local bus PCI DMAs translated into local bus bursts FE/Controls/DAQ Workshop Beat Jost, Cern 12 Other Issues Are the proposed interfaces Controls (I2C, JTAG, parallel bus, (PCI)) Interface

acceptable and sufficient? Hardware configuration, e.g. amount of memory required Cost Interference with on-board analog electronics Reset of controls interface while taking data (SUE recovery) Final Implementation: Glue logic on separate board or integrated on motherboard? FE/Controls/DAQ Workshop Beat Jost, Cern I2C Parallel Bus PCI Bus JTAG Glue Logic Credit-Card PC 13 Pros and Cons Pros only point-to-point connections flexible concerning association between controls WS and electronics board (scalability) programmable element on board allows extensive self-test and diagnostics allows to use (design) cheap crates test-benches and repair stands are very simple and cheap: just

need a PC or terminal and an Ethernet cable and a power supply Cons potentially cost (even though it seems that e.g. smartModules are competitive with VME) difficult to mix in one crate commercial (VME-based?) modules currently only few vendors (issue is more the pin-out), however market is growing, Effort going on to standardize pin-out. FE/Controls/DAQ Workshop Beat Jost, Cern 14 Conclusions We are studying alternatives to crate-based solutions (VME) to be used for the control, configuration and monitoring of board-level electronics Our Goal is to find a solution that is, compared to VME, more flexible more reliable not more expensive (actually much cheaper if cheap crates are used) Credit-Card PCs could be a viable solution The results of the evaluation board will make us know more FE/Controls/DAQ Workshop Beat Jost, Cern 15

smartModule vs. VME Cost Analysis SmartModule Cost (Elan based) VME Based System Cost Component cost [CHF] Unit Cost VME Slave Interface Chip (TUNDRA Trooper II) VME Processor smartModule Glue Logic Ethernet HUB 250 80 30 T otal per Board e.g. PLX 9080 per port Subtotal Glue Logic 360 FE/Controls/DAQ Workshop 60$

5000 Average Per Slot Cost [CHF] 100 333 Quantity of 500 assuming 15 slots occupied (in average) 433 17 450 Beat Jost, Cern 16

Recently Viewed Presentations