GOSSIP : a vertex detector combining a thin gas layer as signal generator with a CMOS readout pixel array Gas On Slimmed SIlicon Pixels.

GOSSIP: a vertex detector combining athin gas layer as signal generator with aCMOS readout pixel array

Gas On Slimmed SIlicon Pixels

Time Projection Chamber (TPC): 2D/3D Drift ChamberThe Ultimate Wire (drift) Chamber

E-field(and B-field)

Wire Plane+Readout Pads

track ofchargedparticle

Wire plane

Pad plane

Let us eliminate wires: wireless wire chambers

1996: F. Sauli: Gas Electron Multiplier (GEM)

1995 Giomataris & Charpak: MicroMegas

Ideally: a preamp/shaper/discriminator channel below each hole….

The MediPix2 pixel CMOS chip

256 x 256 pixelspixel: 55 x 55 μm2

per pixel: - preamp- shaper- 2 discr.- Thresh. DAQ- 14 bit counter

- enable counting- stop counting- readout image frame- reset

We apply the ‘naked’ MediPix2 chipwithout X-ray convertor!

MediPix2 pixel sensorBrass spacer blockPrinted circuit boardAluminum base plate

Micromegas: - 350 V

Cathode (drift) plane: - 700 V

Baseplate

Drift space: 15 mm(gas filled)

Very strong E-field above (CMOS) MediPix!

Cubic drift volume:14 x 14 x 14 mm3

cosmic muon

We always knew, but never saw: the conversion of 55Fe quanta in Ar gas

No source, 1sNo source, 1s5555Fe, 1sFe, 1s

5555Fe, 10sFe, 10s

Friday 13 (!) Feb 2004: signals from a 55Fe source (220 e- per photon); 300 m x 500 m clouds as expected

14 mm

The Medipix CMOS chip facesan electric field of 350 V/50 μm

= 7 kV/mm !!

Eff = e-Thr/G

Thr: threshold setting (#e-)G: Gas amplification

Single electron efficiency

0.00

0.20

0.40

0.60

0.80

1.00

0 1000 2000 3000 4000

Threshold setting (number of electrons)

Eff

icie

ncy

(-)

G=500

G=1000

g=2000

g=4000

g=8000

Expon. (G=500)

Expon. (G=1000)

Expon. (g=2000)

Expon. (g=4000)

Expon. (g=8000)

single-electron avalanche distribution

0

0.0005

0.001

0.0015

0.002

0 1000 2000 3000 4000

electrons in avalanche

Pro

b(n

)

G=500

G=1000

G=2000

G=4000

G=8000

Expon. (G=500)

Expon. (G=1000)

Expon. (G=2000)

Expon. (G=4000)

Expon. (G=8000)Prob(n) = 1/G . e-n/G

• no attachment• homogeneous field in avalanche gap• low gas gain• simple exponential grown of avalanche

No Curran or Polyadistributions but simply:

Single electron efficiency

New trial: NIKHEF, March 30 – April 2, 2004Essential: try to see single electrons from cosmic muons (MIPs)

Pixel preamp threshold: 3000 e- (due to analog-digital X-talk)Required gain: 5000 – 10.000

New MedipixNew Micromegas

Gas: He/Isobutane 80/20 !Gain up to 30 k!He/CF4 80/20

…… It Works!

He/Isobutane80/20Modified MediPix

Sensitive area:14 x 14 x 15 mm3

Drift direction:Verticalmax = 15 mm

He/Isobutane80/20Modified MediPix

Sensitive area:14 x 14 x 15 mm3

Drift direction:Verticalmax = 15 mm

He/Isobutane80/20Modified MediPix

δ-ray?

Sensitive area:14 x 14 x 15 mm3

Drift direction:Verticalmax = 15 mm

MediPix modified by MESA+, Univ. of Twente, The Netherlands

Pixel Pitch: 55 x 55 μm2

Bump Bond pad: 25 μm octagonal75 % surface: passivation Si3N4

New Pixel Pad: 45 x 45 μm2

Insulating surface was 75 %Reduced to 20 %

Non Modified Modified

Vernier, Moire, Nonius effect

Pitch MediPix: 55 μmPitch Micromegas: 60 μm

Periodic variation in gain per 12 pixels

Focussing on (small) anode padContinues anode plane is NOT requiredReduction of source capacity!

Non-modified MediPixModified MediPix has much less Moire effect

No charge spread over2 or 4 pixels

Modified

Non Modified

InGrid: perfect alignment of pixels and grid holes!Small pad: small capacitance!

De-focussing

focusing

focusing

De-focussing

INtegrate Micromegas GRID and pixel sensor

‘Micromegas’

By ‘wafer post processing’at MESA+, Univ. of Twente

InGrid

Integrate GEM/Micromegas and pixel sensor: InGrid

‘GEM’ ‘Micromegas’

By ‘wafer post processing’

-For KABES II, there are two options. The TPC with transverse drift option would need strips rather than pixels.But it could be interesting to have an InGrid-like integrated mesh.The thin Si or CMOS+gas option would need a very high rate capability. -CAST (CERN Axion Solar Telescope) seems to be a more straightforward application.It simply requires a possibility of triggering a common stop.This is why Esther Ferrer-Ribas, from CAST, will join us. - The polarimetry application (challenging Belazzini) is very interestingfor people from the Astrophysics division. The requirement is very similar to CAST's. - The MicroTPC might have applications in nuclear physics or in Babar, for instance. - There are other applications (X-ray beam monitor for SOLEIL) which I can talk about tomorrow. - The protection issue is essential in all Micromegas applications.

! With 1 mm layer of (Ar/Isobutane) gas we have a fast TPC!

• thick enough for 99 % MIP detection efficiency• thin enough for max. drift time < 25 ns (LHC bunchX)

Replace {Si sensor + amplifier} by gas layer:

tracker for intense radiation environment

After all: until 1990 most vertex detectors were gas detectors!Si solved granularity problems associated with wires.

CMOS pixel array

MIP

Micromegas (InGrid)

GOSSIP: Gas On Slimmed SIlicon Pixels

Drift gap: 1 mmMax drift time: 16 ns

MIP

CMOS pixel chip

Cathode foil

Essentials of GOSSIP:

• Generate charge signal in gas instead of Si (e-/ions versus e-/holes)• Amplify # electrons in gas (electron avalanche versus FET preamps)

Then:• No radiation damage in depletion layer or pixel preamp FETs• No power dissipation of preamps, required for Si charge signals• No detector bias current

1 mm gas layer + 20 μm gain gap + CMOS (almost digital!) chipAfter all: it is a TPC with 1 mm drift length (parallax error!)

Max. drift length: 1 mmMax. drift time: 16 nsResolution: 0.1 mm 1.6 ns

AgeingEfficiencyPosition resolutionRate effectsRadiation hardnessHV breakdownsPower dissipationMaterial budget

AgeingRemember the MSGCs……

Little ageing:

• the ratio (anode surface)/(gas volume) is very high w.r.t. wire chambers

• little gas gain: 5 k for GOSSIP, 20 – 200 k for wire chambers

• homogeneous drift field + homogeneous multiplication field versus 1/R field of wire. Absence of high E-field close to a wire: no high electron energy; little production of chemical radicals

Confirmed by measurements (Alfonsi, Colas)

But: critical issue: ageing studies can not be much accelerated!

Efficiency

Determined by gas layer thickness and gas mixture:

Number of clusters per mm: 3 (Ar) – 10 (Isobutane)Number of electrons per cluster: 3 (Ar) - 15 (Isobutane)Probability to have min. 1 cluster in 1 mm Ar: 0.95

With nice gas: eff ~ 0.99 in 1 mm thick layer should be possible

But…….

• Parallax error due to 1 mm thick layer, with 3rd coordinate 0.1 mm:• TPC/ max drift time 16 ns; σ = 0.1 mm; σ = 1.6 ns: feasible!• Lorentz angle

• We want fast drifting ions (rate effect)• little UV photon induced avalanches: good quenching gas

Position resolutionTransversal coordinates limited by:• Diffusion: single electron diffusion 0 – 40/70 µm

weighed fit: ava 20/30 µm10 e- per track: σ = 8/10 µm

• pixel dimensions: 20 x 20 – 50 x 50 μm2 Note: we MUST have sq. pixels: no strips (pad capacity/noise)Good resolution in non-bending plane!Pixel number has NO cost consequence (m2 Si counts)Pixel number has some effect on CMOS power dissipation

• δ-rays: can be recognised & eliminated

3rd (drift) coordinatelimited by:• Pulse height fluctuation• gas gain (5 k), pad capacity, # e- per cluster With Time Over Threshold: σ = 1 ns ~~ 0.1 mm

0

Q

20 ns

Rate effects

time

0

Q

20 ns

• ~10 e- per track (average)• gas gain 5 k• most ions are discharged at grid after traveling time of 20 ns• a few percent enter the drift space:

SLHC @ 2 cm from beam pipe:10 tracks cm-2 25 ns-1

400 MHz cm-2!

Some ions crossing drift space: takes 20 – 200 μs!•ion space charge has NO effect on gas gain• ion charge may influence drift field, but this does little harm• ion charge may influence drift direction: change in lorentz angle ~0.1 rad• B-field should help

Data rate

Hit Pixel (single electron) data: 8 bit column ID8 bit row ID4 bit timing leading edge4 bit timing trailing edge

total 24 bits/hit pixel

100 e-/ 25 ns cm2 380 Gb/s per chip (2 x 2 cm2)

Cluster finding: reduction factor 10: 40 Gb/s

Horisberger:

Data rate, DAQ, data transmission is a limiting factor for SLHC

Required: rad hard optical links with 1 mm3 light emittors per chip!

Radiation hardness

• Gas is refreshed: no damage• CMOS 130 nm technology: ? TID

? NIEL? SEU: design/test

• need only modest pixel input stage• How is 40 Gb/s hit pixel data transferred?

need rad hard optical link per chip!

HV breakdowns: InGrid issue

4) Protection Network

1) High-resistive layer

2) High-resistive layer

3) ‘massive’ pads

Power dissipation

For GOSSIP CMOS Pixel chip:

Per pixel: - input stage (1.8 μA/pixel)- monostable disc/gate

Futher: data transfer logic

guess: 0.1 W/cm2

Gas Cooling feasible!

Detector Material budget

‘Slimmed’ Si CMOS chip: 20 μm SiPixel resistive layer 1 μm SU8 eq.Anode pads 1 μm AlGrid 1 μm AlGrid resistive layer 5 μm SU8 eq.Cathode 1 μm Al

Gas instead of SiPro:- no radiation damage in sensor- modest pixel input circuitry- no bias current, no dark current (in absence of HV breakdowns..!)- requires (almost) only digital CMOS readout chip- low detector material budget

Typical: Si foil. New mechanical concepts: self-supporting pressurised co-centric balloons

- low power dissipation- (12”) CMOS wafer Wafer Post Processing dicing 12” pcs

- no bump bonding- ‘simple’ assembly

- operates at room temperature- less sensitive for X-ray background- 3D track info per layer

Con:- Gas chamber ageing: not known at this stage- Needs gas flow (but can be used for cooling….)

How to proceed?

- InGrid 1 available for tests in October:- rate effects (all except change in drift direction)- ageing (start of test)

Proof-of-principle of signal generator: Xmas 2004!

- InGrid 2: HV breakdowns, beamtests with MediPix (TimePix1 in 2005)

- TimePix2: CMOS chip for Multi Project Wafer test chip

GOSSIPO !

Dummy wafer

Essential Ingredients of GOSSIP CMOS chip

RATE

Assume application in Super LHC:

- Bunch crossing 25 ns- 10 tracks per (25 ns cm2) - 10 e- per track (average: Landau fluct.)

So: 4 MHz/mm2 tracks!, 40 MHz/mm2 single electrons!

Charge signal on pixel input pad

Q

10 - 20 ns

- Signal shape is well defined and uniform- No bias current, no dark current- Signal is subject to exponential distribution- may be large, but limited by

- chamber ageing- space charge (rate) effects

single-electron avalanche distribution

0

0.0005

0.001

0.0015

0.002

0 1000 2000 3000 4000

electrons in avalanche

Pro

b(n

)

G=500

G=1000

G=2000

G=4000

G=8000

Expon. (G=500)

Expon. (G=1000)

Expon. (G=2000)

Expon. (G=4000)

Expon. (G=8000)

Input Pad capacity

preamp stage, noise, power

- Input pads may be small: focusingToo small pads: chamber ageing

- capacity to neighbors & metal layers- capacity due to gas gain grid- Pixel size: 50 x 50 - 20 x 20 μm2

4 fF seems feasible

Time resolution

preamp-disc speed, noise, power

- Measurement 3rd coordinate: σdrift time: 25/16 = 1.5 ns- Time over threshold: slewing correction- drift time related to BX

Record: leading edge - BXtrailing edge - BXBX ID

Data Readout

ALL data: 80 MB s-1 mm-2 ( 15 GB/s per chip)Maybe possible in 10 years from now:

- optical fibre per chip- Vertex can be used as trigger

For SLHC:Use BX ID info (typical Vertex policy)

- tell BX ID to all (Rows/Columns/Pixels)- get data from (Row/Column/Pixel)

Gossipo

MultiProjectWafer submit in 130 nm CMOS technology

Test of essential GOSSIP ingredients:

- Low power, low input capacity preamp/shaper/discriminator- 1.5 ns TDC (per discriminator output)- Data transfer

Maybe not all of this in a first submitMaybe with less ambitious specifications

Amplifier-shaper-discriminator

- How to apply a test pulse?- using gas gain grid (all channels fire)- capacitive coupling test pulse strip- reality: with a gas gain grid(!)

- What to do with the output?- (bonded) contact: digital feedback?!- TDC + DAQ?

TDC

- 1.5 ns clock: derived on-board from 40 MHz BX clock?- 640 MHz clock distribution (per pixel?!)- DLL?

(My) goal of this meeting:

- Are there any showstoppers in this stage?- can we define a Gossipo concept (block diagram)?- Can we estimate the amount of work?