Cooper

Off-Axis Detector with RPCs(Resistive Plate Chambers)John Cooper, Fermilab: 

Off-Axis Detector with RPCs (Resistive Plate Chambers) John Cooper, Fermilab We have an alternate RPC design for the detector It is by no means a final design Plenty of room for new collaborators to help design a better version I believe “better” means “cheaper, but with adequate performance to accomplish the physics goals” This is a fact of life when you talk about building 50 kilotons of anything If it’s too expensive we won’t get to build it at all, or We will have to wait additional years to accumulate the funds to build it I would be happy to trade years of waiting for a less than perfect but adequate detector This talk describes the current RPC version Then suggests avenues of R&D to make it cheaper

The basic RPC unit: 

The basic RPC unit Glass RPCs are our design baseline This is a conservative choice based on the successful BELLE experience with their barrel and endcap muon systems BELLE has 5,000 m2 of such chambers BELLE has operated them without problems for 5 years Signal pickup strips in x Resistive paint Glass plates 8 kV Optional Signal pickup strips in y Resistive paint Spacers ground plane insulator

BELLE design detail – a very simple device: 

BELLE design detail – a very simple device Two large sheets of float glass, 2 mm thick, 1012 Wcm resistivity Noryl spacers every 10 cm, Epoxy 3M 2216 India ink (30%black + 70%white), of order 1 MW per square Gas connectors, Gas: 30% Argon, 62% HFC-134A tetrafluoroethane, 8% Butane, (not flammable) Edge spacers BELLE Experience in 5 years of operation: No degradation in efficiency No chamber exchanged or replaced.

Off-Axis scheme: 

Off-Axis scheme 6 separate chambers, each 2.84m by 2.43m, 3mm thick glass (vs. BELLE 2mm) Arranged in two layers to get full efficiency (offset dead spacer areas) Call this unit an “RRA” , RPC Readout board Assembly 4 layers of glass, 2 layers of strips on Particle Board, ground plane on opposite side, assorted resistive paint and insulating layers 2 more outside Particle Boards for protection RPC double layer Horizontal strips About 4 cm wide, 17 micron Cu foil on Particle Board 64 channels Vertical strips, About 4 cm wide, Cu on Particle Board 3 x 64 = 192 channels Absorber Particle board 8 ft x 28 ft “magic” max size From industry, 1 inch thick 0.7 gm/cc Absorber Particle board If BOTH strips, called XANDY If one per layer, XORY (and half the electronics) Ground plane

Gas plumbing details on the RRAtriangular cutout on glass corners for gas manifold, recirculating gas system with 1 volume change per day: 

Gas flow spacers introduce a 0.92% dead space (but can offset in the two layers, so count as 0.0 %) Edge spacers introduce 0.84 % dead space Gas plumbing details on the RRA triangular cutout on glass corners for gas manifold, recirculating gas system with 1 volume change per day Triangles introduce only 0.19 % dead space Spacers every 20 cm (since 3 mm thick glass)

RRA End View: 

RRA End View 1.27 cm (0.5”) Shelf at bottom Only 0.46 % dead space

Next, assemble 12 RRCs into a module called a “Toaster”: 

Next, assemble 12 RRCs into a module called a “Toaster” Composite particle board corner post formed by sandwiching 3 inch of absorber between two 1/2 inch thick aluminum plates 1/8 in. skin with 1/8 in ribs. Endframe is aluminum on one end, steel on the other end 12 in Structural angle “shelf” for RPC and absorber support. aluminum endwall uses L 6 x 6 x 1/2 Steel end is 3/8 8 rows of 5/8 in countersunk bolts, 3 bolts/row. Metal/particle board surfaces are also glued STRONG structural sidewall particle board absorber Corner Posts inspired by ISO Shipping Containers bear the Toaster weight, Particle Board is self-supporting between the 2 endwalls

Embed Corner fittings on the posts to allow stacking of several modules: 

Embed Corner fittings on the posts to allow stacking of several modules Aluminum Corner Fittings are embedded in the top and bottom 4 inches of composite column (can remove “nipple” for lifting fixture attach point) bottom corner fitting on module above top corner fitting goes here 4 in 3 in Allow 0.5 inch clearance between particle boards in adjacent vertical stack Only 0.46 % dead space

Next, Attach RRC Modules to Ribs: 

Next, Attach RRC Modules to Ribs Six 1/2 in. screws, ¾ in long, 5 inch vertical spacing. Identical pattern is used at bottom Weld Stress and Angle Deflection in Endframes 0.029 inch deflection aluminum steel 0.016 inch max stress = 6 ksi max stress = 9 ksi

Assemblyorder for 12 RRAs and 2 Sidewalls: 

Assembly order for 12 RRAs and 2 Sidewalls

And you have a 24 ton “Toaster” with empty slots: 

And you have a 24 ton “Toaster” with empty slots 1/8 in. skin (ribs not visible) composite aluminum and particle board corner post 4.5 inch RPC Modules attached to ribs Thinner 3.5 inch RPC Modules at sidewalls only

Fill it with Toast (at the Far Site) and you get a 44 ton assembly: 

Fill it with Toast (at the Far Site) and you get a 44 ton assembly All the non-structural particle board absorbers sit on the endwall ledges and are self supporting 4 inch thick absorbers 4.5 inch thick RPC modules

Stack the full Toasters2 wide, 8 high, 75 deep: 

Stack the full Toasters 2 wide, 8 high, 75 deep steel endframes to the outside (cheaper) aluminum endframes at the center crack, Only 0.033% dead space 1200 Toasters in all 70 more rows Total dead space is 1.98 %, Dominated by the Noryl edge spacers on each RPC Minimizing this was a design goal

Slide14: 

HV System sits on the outside edge of the RRAs Cockroft-Walton HV supplies with current readback Implementation for Neutrino Detector 6 C-W supplies mounted on one PC-board One PC-board services 6 RPCs in a double layer Each C-W generates up to + 4,500 V and – 4,500 V Serial control via CANbus CANbus node serves the 12 C-W boards on a module Multiplexed DAC and ADC reside on CANbus node One serial cable and one low voltage cable are all that is needed to provide HV and monitor current to a module

Slide15: 

Electronics is all on the outside edge Discriminate Hits from Detector Timestamp Hits in Front End Store Timestamps in Local Buffers Read Buffers Periodically Use Back End Trigger Processor to Reconstruct Hits System Overview Trigger-less – Like MINOS Similar to a Parallel Development for the Linear Collider Custom ASIC design at Fermilab in collaboration with Argonne LC Primary Goal: Cheap Electronics, 1 Bit Dynamic Range

So I hope you are convinced that we have a complete RPC design for costing and simulation purposes: 

So I hope you are convinced that we have a complete RPC design for costing and simulation purposes As you will hear, this design seems to be more expensive than the basic scintillator design Many parts to build and assemble More parts implies more cost but does give each institution a part of the detector to build Is this the final word? NO, many options remain to be investigated

Some options are simple, not too controversial: 

Some options are simple, not too controversial Change the gas system from copper to some plastic tubing impervious to water? BELLE had an initial problem with water vapor through polyflow PVDF tubing may work Combine the HV and gas systems for each of the two RPC layers in an RRA Lose ability to control separately, but these chambers are robust, so why have the unused extra control? Simplify the gas system manifold? Remove scintered metal strainer and flow restrictor (0.25 mm I.D.) per triplet of RPCs – lose perfectly balanced flow to all RPCs Reduce gas flow, don’t recirculate? We have 1 volume change per day, BELLE did per 2 days Maybe can flow at 0.1 – 0.2 per day in pulsed flow with a long output tube open to atmosphere? (OK, this one is pretty controversial)

Some options are more ambitious: 

Some options are more ambitious Drop double RPC layer Single layer is 93-94 % efficient including dead area from spacers Compensate by operating RPCs in avalanche mode instead of streamer mode Increases efficiency per layer to 98-99 %, so one layer may be enough Signals are 100 times smaller 200 mV into 100 ohm  2 mV into 100 ohm Avalanche Electronics design is in common with US LC hadron calorimeter groups (Jose Repond at ANL, Ray Yarema at Fermilab) so we can test this Alternate scheme, drop double layer, stay in streamer mode, but add 5% more layers to compensate for 5% drop in efficiency Replace the Copper strips with Aluminum Strips already laminated ( $0.15 + $0.30/sq ft  $0.20 or less) The original scheme used Johns-Manville AP Foil-Faced polyisocynurate foam sheathing or DOW THERMAX foam board plus a kraft paper / aluminum laminate (another std building material) Could not find an reliable cheap way to attach cables to Aluminum strips Could we deliberately do a capacitive coupled connection, using a controlled thickness spacer with copper tape overlay at the ends of strips?

Some options are more radical: 

Some options are more radical The present design was a compromise between RPC enthusiasts in favor of monolithic structures (like the liquid scintillator) and modular structures using intermodal shipping containers We compromised so that “a solution” could be written up by our small group of 4 physicists and 6 engineers on a deadline of last September In retrospect we all agree we compromised on a solution that has the world’s most expensive custom container, a container so large that we can’t fill it and transport it on US Interstate highways. This is why we build a toaster but add the toast at the far site Committee design of an elephant resulted in a white elephant? So we are again separately pursuing the monolith & container solutions Each seeking a lower cost with adequate performance We still meet together every week

RPCs in a monolithic design – Options: 

RPCs in a monolithic design – Options Overall the monolithic design, completely assembled at the far site may be cheaper At least the assembly should be as cheap as the very similar Liquid Scintillator assembly This can reduce the number of vertical strip readout channels by 50% 8 foot high vertical strips become 28 ft (or even 56 ft high?) 8 vertical strip channels become 2 channels (or even 1 channel?)

RPCs in a shipping container design -- Options: 

RPCs in a shipping container design -- Options 1200 custom 28 ft long modules replaced with 2400 standard 20 ft steel shipping containers $ 3500 custom module becomes $1000(used) or $1500(new) container Readily available throughout the U.S. (trade imbalance) Full modules can be built and tested at many sites Nice for collaboration, but have to watch transportation costs Use the corner fittings to lift the containers with a crane

More on RPCs in Containers: 

More on RPCs in Containers A steel container does not need to rely on the strength of particle board absorber since the container has a fully supported floor Price of particle board is $ 0.13 per pound and we need $ 12.4 M of it Radiation length is 53.6 cm, density is 0.65 gm/cc An alternate material is Drywall or Sheetrock This is Gypsum, Ca SO4 2H20, Calcium Sulfate Dihydrate US annual output is 38,000 kilotons, Cost is about $ 0.05 per pound Radiation length is 37.9 cm, density is 0.68 gm/cc Another alternate is Cellular Foam Concrete This is Portland cement, sand, water, and “shaving cream” 50% Tricalcium Silicate, 25% Dicalcium Silicate, 10% Tricalcium Aluminate, 10%Tetracalcium Aluminoferrite, all hydrated, 5% gypsum, Not common in the US, invented in Europe, US price about $ 0.10 per lb Radiation length can be made at 47 cm, density of 0.7 gm/cc with sand : portland cement at 3.5 : 1, i.e “structural sand” Pour in place, labor savings?

More on RPCs in Containers: 

More on RPCs in Containers Standard ISO containers introduce a new problem there is 20 cm high dead space or “crack” at the bottom of the container for fork lift pockets and the container floor and another dead space at the door end of the container So 2 % dead area  1.9 +7.6 + 2.6 % = 12.1 % dead area Can software for a “Tracking” Calorimeter keep track of where each track passes through such cracks and compensate? ISO containers at the US Interstate weight limit can stack as high as 11 on 1 (31 m or 100 ft high) The detector building appears to be significantly cheaper (at equal volume) if higher but with a smaller area footprint

Slide24: 

A Containerized Detector is mobile We may not know the optimum site early? We may want to move to a different site after seeing a signal? 7,500 TEU Ship 20 ft in a C-130 Double Stack Rail car 20 ft on a truck chassis

Slide25: 

Intermodal Landbridge by Rail & the cost to ship a 20-foot container $1066 $ 867 $706 $800 LA + Long Beach 9.6 M (8.2 M in 1999) NY, NJ 3.3 M Houston 1.0 M Chicago Rail 8.8 M in 1998 WEST Burlington Northern Santa Fe or Union Pacific EAST CSX or Norfolk Southern Charleston + Virginia + Savanna 3.8 M Seattle + Tacoma + Vancouver 3.7 M TEU in 2002 I think this explains why the cost of containers is fairly uniform across the US Costs from BNSF website