1. 18 Sep 2008Paul Dauncey 1 DECAL: Motivation Hence, number of charged particles is an intrinsically better measure than the energy deposited Clearest with ideal calorimeter; no experimental effects Energy deposited (“analogue” ECAL) resolution ~50% worse than number of particles (“digital” ECAL) resolution Average number of charged particles in an EM shower  incident energy Fluctuations around the average occur due to statistical nature of the shower Average energy deposited in the sensitive layers  number of charged particles Fluctuations around the average occur due to angle of incidence, velocity and Landau spread 20×0.6X 0 + 10×1.2X 0 Sensitive Layers

2. 18 Sep 2008Paul Dauncey 2 Digital ECAL concept Can we measure the number of charged particles directly? It is possible to get close to the analogue ideal resolution with low noise electronics Can we get anywhere near the ideal resolution for the digital case? Make pixellated detector with small pixels Probability of more than one charged particle per pixel must be small Allows binary readout = hit/no hit EM shower density ~100/mm 2 in core so need pixels ~ 50  m Results in huge number of pixels in a real ECAL ~ 10 12 pixels Need readout integrated into pixel Implement as CMOS MAPS sensor

3. 18 Sep 2008Paul Dauncey 3 MIP counting Cannot simply count hit pixels Charge diffusion can cause nearest neighbours to fire too Fluctuations in neighbours would dominate resolution Need to do neighbouring hit “clustering” to convert hits to particle count Called “MIP counting”; used for all of the following Hits in a layer before (left) and after charge diffusion (right)

4. 18 Sep 2008Paul Dauncey 4 Critical parameters for resolution Degradation from ideal resolution arises from Noise hits Dead area Charge diffusion to neighbouring pixels Particles crossing pixels boundaries and sharing pixels Importance of various effects differs Illustrate with 10GeV showers Noise ~ 5% Diffusion ~ 5% Dead area ~ 5-10%

5. 18 Sep 2008Paul Dauncey 5 Effect of hit confusion “MIP counting” helps but it is not clear where the limit is How much is due to the algorithm? How much is due to the shower? Basic properties of an EM shower not verified at 50  m Which algorithm to use depends on shower details which may not be modelled well Currently gives remaining ~ 40% degradation to resolution so this is by far the dominant effect Essential to get experimental data on fine structure of showers to know the ultimate realistic resolution of a DECAL Major study of MIP counting algorithms must be done on these data

6. 23 Jul 2008Paul Dauncey 6 Overall aims of DECAL Workpackage 4 Developing a technique not a technology Aim to do an existence proof of a DECAL Doing a calorimeter experiment using CMOS sensors Not trying to design best CMOS sensor for a DECAL right now Need something which is “good enough” for the test Which effect is most important in limiting the resolution? Simulation says hit confusion, which depends on real shower properties Need real data to confirm this and to test improved MIP counting algorithms Need to check if this is true; limit may in reality be set by other factors Efficiency per pixel? Dead areas? Signal size or S/N? Charge diffusion between pixels? Pixel-to-pixel variations? Incorrect EM shower physics; e.g. simulation wrong for low energy photons? Aims of the workpackage are To see whether a DECAL works at all Check if performance agrees with expectations Get operational and analysis experience Find out where the limitations are so a better sensor could be designed in future

7. 18 Sep 2008Paul Dauncey 7 Workpackage 4: DECAL stack Use TPAC2 sensor tested in WP3 to make a DECAL stack 450×450 ~ 200k pixels and 2.5×2.5cm 2 ; a factor ten in area; otherwise a scale-up of existing sensor 5×5 sensors = 12.5×12.5cm2 per PCB; 5M pixels in a layer 16 layers to make a Si-W DECAL stack; 80M pixels total Sufficient for DECAL proof-of-principle To pack sensors in the plane, will wirebond through slots in PCB Aim for pixel-pixel gap between sensors to be only 500  m ~ 4% extra dead area “Real” detector would bump-bond but we wanted to minimise engineering effort/risk Gives “packing fraction”, i.e. dead area, close to realistic DECAL

8. 18 Sep 2008Paul Dauncey 8 Workpackage 4: Tasks TPAC2 evaluated and tested in WP3 by mid 2010 Sensor testing and stack assembly: mid-end 2010 PCB and DAQ design (Birmingham, Imperial) Sensor probing and QA (Imperial) Sensor mounting and PCB QA (RAL) Cosmics stack and system tests (Birmingham) Exposure to test beam: begin-mid 2011 Around four weeks of beam time ~30Hz DAQ rate, ~10M events total, ~1TByte Electrons with energies from 1-50 GeV Several stack geometries Pions and muons for calibration/alignment Analysis and simulation: mid 2011 to end of grant Studies of DECAL resolution Modelling of sensor response Comparison to simulation

9. 18 Sep 2008Paul Dauncey 9 Backup: Effect of noise Noise adds hits to showers so increases  N Depends very strongly on threshold Need to increase threshold above noise “wall” Noise has no effect for higher thresholds Resolution degradation ~ 5% If S/N can be improved, then get a plateau so noise has no effect at all on resolution

10. 18 Sep 2008Paul Dauncey 10 Backup: Effect of dead area Existing test sensor has 11% dead region due to on-pixel memory Bands of 250  m wide spaced every 2.4mm Shower width ~ 1cm so every shower sees several dead bands Always loses 11% of hits with small fluctuations Since  E /E  1/  N, impact is not large Gives 1/  (0.89) ~ 1.06 effect Hence ~ 6% degradation Assumes sensor large enough that edge effects are negligible May add ~ 4% more dead area in reality so ~ 2% more to resolution

11. 18 Sep 2008Paul Dauncey 11 Backup: Effect of charge diffusion Some uncertainty in exact level of charge diffusion Verified experimentally to ~20% Simulation allows charge diffusion to be turned on or off completely Very extreme limits The effect of on/off charge diffusion on the EM resolution is ~ 5% Uncertainty on diffusion is negligible

12. 18 Sep 2008Paul Dauncey 12 Backup: Simulation expectation Current extrapolation to “real” detector shows significant degradation of ideal DECAL resolution 35% increase in error Number of pixels hit not trivially related to number of charged tracks Degradation arises from Noise hits Dead area Charge diffusion to neighbouring pixels Particles crossing pixels boundaries and sharing pixels Importance of various effects differs Illustrate with 10GeV showers in next few slides 10GeV 0.032  0.044

13. 18 Sep 2008Paul Dauncey 13 Backup: DECAL 16-layer stack 16 layers gives degraded resolution by factor ~ 50% compared with realistic number of layers (~30) Compromise of resolution vs. cost Would give essential data on shower hit properties Extrapolate to realistic calorimeter sampling using simulation to get actual DECAL resolution