LAV contribution to the NA62 trigger Mauro Raggi, LNF ONLINE WG CERN 9/2/2011.

LAV contribution to the NA62 trigger

Mauro Raggi, LNF

ONLINE WG

CERN 9/2/2011

LAV in the L0 or L1 trigger?1. LAV can help rejecting photons from p+p0 decay and m2. LAV can provide Time (Dt 500ps), Energy (20%), f

position3. The LAV detector is able to provide all its information

directly @ L0TP4. The LAV it’s able to send all it’s data to L1 PCs

following L0TP request

We have the possibility to choose in which trigger level we use the LAV information

Why the LAV in LVL0?

LAV MIP primitives to L0

Identify MIPs (μ or π) and to distinguish them from photons and electrons:

• N blocks per ring ≤ 2 for each involved ring

• 3 ≤ N firing blocks ≤ 6

• E_bl < 200 MeV for each block over threshold

• E_ring(i) /E_ring(i+1) < 2 for each pair of rings

• N_cl = 1, one cluster (only) in the LAV station

The LAV MIP trigger can be matched with a MUV hit to distinguish pions from muons

LAV MIP efficiency test beam

MIP single block efficiency of the LAV is >95% at the foreseen thr 7mV

We will have >3 blocks per MIP in each LAV so the efficiency will rise well above 99%

Can LAV12 be used, in absence of the CHOD, to compensate the geometrical inefficiency of MUV in covering the RICH triggered muons?

L0 algorithm efficiency to be studied!

EM showers trigger

Not necessary very efficient for very low energy particles @ L0?

All conditions can be used in or or and

• Ntot > 10 OR Etot > 1 GeV

• Ering > 0.1 GeV x Nring for al least 2 rings

• E_ring(i) /E_ring(i+1)> 2 for at least a pair of rings

The real algorithm will be tuned using the LAV MC for L0

Purity for low energy photon is an issue

MC multiplicity MIP showers

Muon 1 GeV

Electron 1 GeV

Not trivial to distinguish low energy photons form muons

P. Massarotti form LAV MC

ANTI-A12

Producing the L0 LAV STEP I

STEP I: only ANTI-A12 in L0:Compute only ANTI-12 primitives and send to L0TPReceive the L0TP response and send all hits to L1 PCsReconstruct the whole detector information in PCs

Producing the L0 LAV STEP IAdvantages of having only ANTI-A12 in the L0

• Only needs firmware development

• The algorithm only involves 1 ANTI

• Don’t require TEL62 communication & LAV data condensation

• Save a lot of man power (good for the synchronization or early run)

Disadvantages

• Lose ~ 50% of LAV rejection capability in L0

• Lose some muon discrimination capability

• All the rejection is recovered in L1

Producing the L0 LAV STEP II

What we need for step IIScenario n 2 all the 12 LAV directly in LvL0 (as in the TDR)• Compute local primitives into single ANTI and send to

concentrator

• Condensate all the 12 station’s primitives into a LAV primitive

• Send the LAV primitive to L0TP and receive the response

• Distribute the L0TP response to all the LAV’s TEL62

• All the TEL62 send the data to L1 PC’s using remaining eth interface

Advantages• Full LAV rejection capability @L0 (50% better rejection p+p0)

• No need of having the LAV in the L1 trigger (only integrity checks)

Disadvantages• Need to project and build 4Gbit Eth receiver for TEL62

mezzanine

• Can a commercial 24x1Gbit Eth switch be enough?

• Need 1 more dedicated TEL62

• Need to develop a dedicated firmware for the concentrator TEL62

ConclusionsLAV is able to send all it’s raw hits to LvL1

• Care should be taken to noisy channels

The LAV can produce 2 primitives @ L0

• MIP primitive (p or m) order 99% efficiency

• Shower primitive (e or g) low E purity to be understood

LAV can have 2 approach to trigger

• Participate to L0 with only ANTI-A12 50% rejectiono Exploit the rest of the rejection at L1

• Use the full detector directly at L0 (all 12 stations)o The data load seems feasible

o The computing required in the TEL62 has to be understood

NA62 trigger rates overview

At which level is the LAV system able to send data?How many data will it send?

LAV expected raw rates TDR

In principle the TEL62 is able to transmit all the hits of the LAV using 3 Gbit eth interface leaving 1 free for the trigger.

From TDR

Using an hit size of 192 bits and only 7 hits per event (MIP)

Simulation of LAV rejection factors

What is already done from Spasimir it’s more or less all

• Assume that the LAV is efficient for all the impinging photons is a good approximation @ trigger level

• The only upgrade could be to implement a thr at 50 MeV below which the detector is inefficient

• Try to understand the L0 LAV algorithm efficiency for low energy gamma.

LAV maximum data volume revised

Trigger LVL Rate Destination

Max data vol.* Total**

Raw LAV rate 11 MHz/12 sta

LvL 0 26Gbit/36 lnk

LVL 0 1 MHz L0 req.

Input LvL 1 2.4 Gbit/station 14.5 Gbit/36 lnk

LVL 1 100KHz L1 req.

Input LvL 2 0.24 Gbit/station

1.5 Gbit/36 lnk

LVL 2 15 KHz L2 req.

disks 4.5 Mbyte/sRaw LAV rate computed assuming:15 hits/station and 160bits per hit (included extra hits in the TDC)Due to the fact that the rate is MIP dominated we have more than a factor 2 safety factor (expected hits per MIP ~7 hits/station)

* Assumes the station has data to send to LVL1 for each LVL0 request (pessimistic)** Assumes 50% of the stations will respond to each LVL0 request (pessimistic)

LAV system should be able to send all its data, without any reduction, to LvL1 following a LvL0 request!

Hit size

Each hit will produce 4 of this word

• 2x32bit leading + 2x32bit trailing

Assuming to add a 32 bits time stamp to the hit

Total hit size is 160bits

Maximum latency in TEL62 DDR

Assuming the highest rate ANTI-A1 1.77MHz

Assuming a factor 2 more hit 15/stat (refirings)

Total size of data produced by a single ANTI:

• 15hits*160bit*2MHz = 4.8 Gbit/s

• 600 Mbyte/s into TEL62 memories

TEL62 has 2Gbyte or 4Gbyte DDR memories

• 2Gbyte means ~ 3.3 s ANTI data

• 4Gbyte means ~ 6.6 s ANTI data

Data load to the concentrator

The concentrator will receive the primitives from the LAV stations with following information

• Trig. type (muon or gamma)

• Event time (Fine time + time stamp) 40 bits

• Event Energy (most probably charge)

• Event position (phi position of the seed in the LAV station)

Ev. type Ev. Energy Ev. Position

• Assuming as rate the LAV total rate of 11MHz and data size 64bit– Total data load to concentrator= 11MHz*64bit = 704 Mbit/s– The rate divided over 12Gbit eth can be managed by 1

TEL62

Data load from concentrator to L0TP

For each local primitive the concentrator should send 2x32bit word to L0TP shown above:

Data to L0TP=2x32x4.5E6 = 288 Mbit/s = 36MByte/s using the LAV OR rate 4.5MHzData to L0TP=2x32x11.5E6= 740 Mbit/s = 93MByte/s using the LAV tot rate 11.5MHz

The concentrator task is to condensate the 12 station information in a single primitive message to L0TPThe concentrator should reduce the total LAV rate 11MHz to the LAV OR rate of 4.5 MHz

Reducing the data volume @ LVL1

Reduce the number of extra hits

• Define in the FPGA an algorithm which suppress extra hit in the same TDC channel

Reduce the size of the hit information

• Transmit only hit charge and time.

Condensate the LAV information

• Provide particle based info instead of hits

Transmit only charge and time

HPTDC produces 4x32 bits word per hit

• 2x32bit leading and 2x32bit trailing edge

• 32bit time stamp

• 160bit per hit

If compute a time and charge in the TEL62 or @ L1

• 1x12 bit charge + 1x8bit fine time + 12 bit channel

• 1x32 bit time stamp

• 64 bit per hit

Can gain more than a factor of 2 in hit size!