federico alessio richard jacobsson

13
TELL40 FW discussion: «upgraded» TFC points of view Federico Alessio Richard Jacobsson

Upload: shaeleigh-aguirre

Post on 31-Dec-2015

19 views

Category:

Documents


3 download

DESCRIPTION

TELL40 FW discussion : « upgraded » TFC points of view. Federico Alessio Richard Jacobsson. Reminder: the logical S-TFC at a glance. Committment to provide : S-ODIN firmware TFC+ECSInterface firmware TFC commands decoding block in TELL40  VHDL entity to be plugged in - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Federico Alessio Richard Jacobsson

TELL40 FW discussion:

«upgraded» TFC points of view

Federico AlessioRichard Jacobsson

Page 2: Federico Alessio Richard Jacobsson

TELL40 FW Workshop, 02/03/12 F. Alessio, R. Jacobsson 2

Reminder: the logical S-TFC at a glance

S-ODIN

TFC+ECS Interfaceto FE / to TELL40s

TELL40s

Front-Ends

LHC Interfaces

TFC

TFC + ECS

TFC

S-EFFECS (FE)

LOW LEVEL TRIGGER

THROTTLEECS

DATA

DATA

Committment to provide:- S-ODIN firmware- TFC+ECSInterface firmware- TFC commands decoding block in

TELL40 VHDL entity to be plugged in Input parallel TFC word, output

commands

Page 3: Federico Alessio Richard Jacobsson

TELL40 FW Workshop, 02/03/12 F. Alessio, R. Jacobsson 3

S-TFC documentation

- TFC system-level specifications note for the Upgrade just published

LHCB-PUB-2012-001 Thanks for all the useful comments, please continue

reading it!

Soon:- Functional specifications (shorter) for BE and FE

detailed definitions and usages of TFC commands in BE and FE

aim at April

Page 4: Federico Alessio Richard Jacobsson

LHCb Data Processing Workshop, 16/11/11 F. Alessio, R. Jacobsson

TFC Back-End Word Format

TFC Word to BE: 44 bits (60 with Reed-Salomon encoder) @ 40 MHz = 1.76 (2.4) Gb/s

THROTTLE Information from BE: 1 bit per board connected to TFC+ECSInterface. Full set of bits sent to S-ODIN by TFC+ECSInterface.

Constant latency after S-ODIN

Encoding 43 .. 32 31 .. 16 15 ..12 11 .. 8

TFC Info BID(11..0) MEP Dest(15..0) Trigger Type(3..0) Calib Type(3..0)

7 6 5 4 3 2 1 0

Trigger BX Veto NZS Header Only BE reset FE reset EID reset BID reset

Page 5: Federico Alessio Richard Jacobsson

LHCb Data Processing Workshop, 16/11/11 F. Alessio, R. Jacobsson

Control functions for Back-End– Same as Front-End

• Bunch ID for synchronization check with internal counter and data from FE• Bunch Counter Reset• Event Counter Reset (reset of same counters as FE + all event related

counters)• Header Only Force FE to transmit only header and no data (Informative)• Calibration pulsing (informative)• Non-zero suppressed readout of current crossing (Informative)• Bunch Crossing Type Veto (Informative)• Front-End electronics reset (Expect only header from FE)

– Back-End Reset (Header Only from FE during reset)– Trigger

• Reject data (Header still sent to farm or not?)• Attention: In TFC word, the trigger (& MEP destination) is not associated to

the transmitted BunchID and the rest of the TFC word• S-ODIN pipes the asynchronous local trigger information for the maximum

latency possible for BE • Realignment of all data for BE is done in TFC+ECSInterface via pipeline

logic– Trigger Type to define type of event, processing type, destination type etc– Multi-Event Packet Destination IP

• Transmitted when MEP should be closed and transmitted– Any other needs? Reserve bits

TFC Back-End control commands

Page 6: Federico Alessio Richard Jacobsson

TELL40 FW Workshop, 02/03/12 F. Alessio, R. Jacobsson 6

General comments

Development of S-ODIN and TFC+ECSInterface firmwares are decoupled from the development of TELL40 firmware:

• Independent VHDL entities with dedicated functionalities• Not many common points in the development (of FW)

«plug&play» TFC decoding block for TELL40 provided as VHDL entity

S-ODIN and TFC+ECSInterface firmwares will be developed as VHDL code which will «plugged in» in their dedicated AMC40 FPGA

i.e., a VHDL block with inputs and outputs to the GX transceivers

Page 7: Federico Alessio Richard Jacobsson

TELL40 FW Workshop, 02/03/12 F. Alessio, R. Jacobsson 7

General comments

But, common points (not from us though..):

• Profit from common low-level interface of AMC40 from Marseille a common way to use the links interface is definitely a plus! collaboration wide validation of the system

• Profit from common ECS-CCPC development… but with some personalized variants … :

S-ODIN and TFC+ECSInterface needs hundreds of monitoring registers

counters and status registers above allAlso, possibility to change registers «on the fly», while running:

we need maximum flexibility at any moment during running! «real time control» of FPGAs … then the VHDL code will protect the readout control

Page 8: Federico Alessio Richard Jacobsson

LHCb Data Processing Workshop, 16/11/11 F. Alessio, R. Jacobsson

Backups

Page 9: Federico Alessio Richard Jacobsson

TELL40 FW Workshop, 02/03/12 F. Alessio, R. Jacobsson

S-TFC concept reminder

9

LHC Timing & Info

Clock Fanout

CLK

TFC SERVER

ECS

PH

Y

Programmable Switch layer (Partitioning)

Built-in GX Transceivers layer

PH

Y FARM

n

n

#links = #LHCb sub-systems ~20m distance

2.4-3.0 Gb/s optical

Super-ODIN

TFC Interface

TF

C, T

hrot

tle

TFC-Master logic

FAN-OUT/FAN-IN Logic

+optional

TFC Master Logic

Switch Logic

Bui

lt-in

S

ER

DE

S la

yer

BE modules

Master FPGA (STRATIX IV GX) Slave FPGA (NIOS II on

CYCLONE III)

TFC-Master Instantiations (x6)

Master FPGA (ARRIA II GX)

FA

N-O

UT

FE interface

}BE modules{

Interaction Trigger

Page 10: Federico Alessio Richard Jacobsson

LHCb Data Processing Workshop, 16/11/11 F. Alessio, R. Jacobsson

Control functions for Front-End– Bunch ID for synchronization check with internal counter– Bunch Counter Reset– Event Counter Reset

• Reset of counter for accepted crossings = crossings for which header+data was sent

• Reset of counter of truncated events• And all other event related counters (TFC command counters, etc!)

– Header Only Force FE to transmit only header and no data– Calibration pulsing (How many types do we need?)– Non-zero suppressed readout of current crossing

• Following n crossing will receive “Header Only” Header only transmission– Bunch Crossing Type Veto based on crossing scheme from LHC

• Send header only for empty crossings and most single beam crossings– Front-End electronics reset

• During the time of the reset (common duration) Front-End receives “Header Only” command and should transmit header only

– Any other needs? Reserve bits All TFC commands (individual signal) require local configurable delay

TFC Front-End control commands

Page 11: Federico Alessio Richard Jacobsson

LHCb Data Processing Workshop, 16/11/11 F. Alessio, R. Jacobsson

TFC Word to FE: 24 bits inside GBT frame @ 40 MHz = ~1 Gb/s

Front-End TFC Word format

Encoding 23 .. 12 11 .. 9 8 .. 5 4 3 2 1 0

TFC Info BID(11..0) Reserve Calib Type(3..0) BX Veto NZS Data Force

FE reset

BID reset

Table 1: Preliminary encoding of the TFC information transmitted to the FE electronics together with ECS

Header Only

All TFC commands (individual signal) require local configurable delays

Page 12: Federico Alessio Richard Jacobsson

LHCb Data Processing Workshop, 16/11/11 F. Alessio, R. JacobssonT

FC

24-

bits

EC

S 5

6-bi

ts

E-LinkProtcolDrivers

GBT Transmitter

TF

C 2

4-bi

tsE

CS

56-

bits

E-LinkProtcolDrivers

GBT Transmitter

PCIe SlavePCIe

MemoryMap

CCPC

TFC Relay & Alignment

GBT-like

TELL40s (44 bits)

S-ODINSynchronous

FE Info fan-out

ECS

FE

TF

C 2

4-bi

tsE

CS

56-

bits

E-LinkProtcolDrivers

GBT Transmitter

Relay/merge block logic:

‘Protocol drivers’ build GBT-SCA packets with addressing scheme and bus type for SCA user busses:- I2C, JTAG, Single-wire, parallel-port,

JTAG, Memory, Temperature sensing, ADC

‘Memory Map’ with internal address scheme for GBT addressing, E-link

addressing and bus type

General Purpose protocol through DIM server on Ethernet (goes PCIe)

TFC+ECS Interface

ECS on “best effort”

Page 13: Federico Alessio Richard Jacobsson

TELL40 FW Workshop, 02/03/12 F. Alessio, R. Jacobsson

S-ODIN data bank

13

S-ODIN transmits a data bank for each accepted event in a MEP

S-ODIN data bank and LLT data bank is merged(reminder: LLT is in same board as new S-ODIN)

Info about timestamp, trigger type, bxid, trigger decision… Mostly like now

Will need at least 10GbE connection directly to FARM what about 40GbE…? has to allow bandwidth partitioning as well

In fact «several» 10GbE (n*10GbE…)

reduced bank size for local tests No LLT for instance