UMICH-2015: Instrumentation II break-out session 1
Goals: There are 3 goals for this session: (1) identify requirements for CMB-S4 telescopes (especially requirements that will be challenging); (2) review the status of current technologies (with a focus on capturing the issues, not on choosing specific approaches); and (3) identify what development/modeling needs to be done for CMB-S4.
Please look at the topics below and come prepared to add quantitative detail, or suggest well-posed questions. If you have a slide to explain your point or question, please post it on the wiki.
What size & frequency telescopes will be needed for CMB-S4?
- Mainly driven by science requirements, but there will be some size vs. performance trade off for larger telescopes
- <1m for inflationary B-modes; neutrino mass, delensing, and other physics require larger telescopes
- Smaller telescopes cost less (cost~diameter^2-3)
- Larger telescopes support more detectors Crossed-Dragone throughput comparison
- Different systematics for large vs. small telescopes T to P leakage for SPT
- Do we want roughly the same beamwidth at all frequencies? build 2 telescope sizes
- CMB-S4 may require multiple telescope sizes.
What are the key optical performance requirements?
- Beam shape and sidelobe level
- EPIC goals for 1nK rms at l=100: E/H gain <1e-4; E/H offset and ellipticity <beam/1000; E/H size <beam/100; <-80dB sidelobes at 20deg (N.B. no ground pickup for EPIC), E/H orthogonality <8', Pixel rotation <3'
- SPT has -100dB sidelobes from panel gaps, giving ground pickup comparable with B-modes for r=0.01
- <1.5” or <beam/100 (from Hu et al. 2003), <beam/28 (EPIC goal for Q to U offset)
Any other challenging telescope requirements?
What is the status of all reflector vs. all refractor vs. mixed designs?
- Existing ground-based experiments use refractors or lens-fed reflectors BICEP/Keck
- Large, broadband, dielectric elements are challenging
- Should we pursue designs with only reflectors?
How will scattering be controlled?
What shielding will be needed?
- Existing small telescopes have comoving absorptive shields BICEP1 forebaffle
- ACT, SPT, Polarbear have comoving reflecting shields
- Shields are generally easier/cheaper for small telescopes CLASS baffles
- Do we need fixed, reflective, ground shields?
- Varying snow buildup can be a problem for large shields
How do we deal with RFI and magnetic pickup? What do we need to measure/model?
Will multiple small telescopes be mounted on a single platform? Can a single mount design be used for all CMB-S4 telescopes?
- Slow rotation comes for free at a mid latitude site
- Devices that modulate polarization are part of the next session
PRINCIPLE THAT CAME OUT OF SESSION
Need to decide on what things you design in to the instrument and what things you decide to correct later. General idea is to look at strawman instrument designs that will require minimal corrections and filtering, and their costs, and then other designs with more corrections/filtering and their costs, etc.
- Feed this question back to the science definition: how much value is there to getting an unfiltered map of the sky (unbiased) -- what spatial modes is it least painful to lose, etc.
- Mike \& Nils will work on doing a common telescope design comparison of multiple optics tubes versus single one.
- Study the question of cost versus broadband per optics tube or telescope.
- Study trade-offs and costs for monolithic designs: materials, quantity discounts, surface finishes, stability.
- Look at the value of fixed ground screens -- compare eg ACT (with screen) in the Atacama to SPT at the south pole (no screen) to PB in the Atacama (no screen).
- Look at the costs of co-moving ground screens in various configurations, and the trade-offs of absorbing vs reflective.
- Come up with machinery to do simulations based on scan strategy and telescope side lobe estimates (including stability of the sidelobes -- to temperature, wind etc, and stability of source when it is the ground rather than the galaxy, eg.)
NOTES TAKEN IN REAL TIME:
- Charles Lawrence slide: if want to match beams at different frequencies, will need different sized telescopes.
- Do we want to match beams, says Clem? Synchrotron has more impact at larger scales than small says Adrian. Can always synthesize a low-angular resolution map out of a high one says Clem; power spectra are pretty red says John C. If doing a map-based subtraction you need matched beams but if doing it as cross-spectra, not required necessarily says Clem.
- It is a loop-back says Clem, systematics control may push you to smaller telescope even though the science requirements might only require larger telescopes on the surface.
- Jeff McMahon's slide; can get good control of main beam effects with bigger telescopes and get them out of the important $\elll$ ranges.
- Clem -- but, with large telescopes will be hard to obtain good far sidelobes because won't have co-moving absorbing forebaffle. Note that QUIET was 2m telescope that did have one.
- Desirable qualities for far sidelobes would be rotation of the instrument around the boresight, large co-moving baffle.
- Lyman -- what is the far sidelobe level in dBi, which is what is relevant, for BICEP? (See later plot posted by Ki Won in dB.)
- Mike Niemack's slide: higher throughput for larger telescopes.
- ACTPol is using about half the achievable throughput of the diffraction-limited FOV, and in fact the correctable (with multiple lens tubes) FOV is bigger than the diffraction-limited one. Crossed Dragone is great, but so is faster optics, lower F\#.
- Mike shows a 5m design with a folding flat to make it compact that can illuminate $10^5$ detectors using re-imaging optics (each hexagon is a lens). Lose only 10\% of the focal plane area for supports for separate reimaging optics. Would need large baffle to protect secondary and primary.
- Steve -- will the good FOV (low aberrations) improvement with F\# hold up once you put in the optics tubes? Mike -- I want to confirm but I think so based on my experience. Steve -- tried things like this for CCAT but when you didn't do the re-imaging of the whole field at once with large lenses etcetera, broke up some of the trends.
- Clem is worried about how close the rays come to the secondary in the design which is not yet optimized; Steve shows EPIC design, which is similar, with scattering off the top of the secondary causing the brightest red spot in plane to the right of the main beam.
- Mike -- critical difference cf POLAR array and EPIC here is having a cold image of the primary in the lens tubes. Akito -- true for the primary but not the secondary. QUIET got away with warm absorbing baffle but harder for bolometers.
- Clem -- if want really high angular resolution, telescopes will be very expensive, and so in that case the design (ie one like this) that packs the largest number of detectors in there might be the best cost trade-off.
- Mike listed specific action items for a trade-off study between the too concepts (small optics tube vs large for a common telescope design, eg.)
- ACTION ITEM: Mike \& Nils will work on doing that common telescope design comparison.
- Akito -- how much of the cost of the telescope depends on the mass? Because CD is a lot heavier vs the Gregorian, because each mirror is large. Steve claims it is about \$20/kg. John C says usually you do end up with expensive mounts for Gregorians because have counterweights.
- Steve -- should we go back to all-reflector designs? Ie, no lenses anywhere. If we are starting from scratch.
- Scaling of cost to diameter it is something like $d^2.5$ but at NRAO for longer wavelengths they use $d^2.7$ and for submm where need smoother surfaces may be $d^3$.
- Nils -- a concrete question is how broadband does a given telescope (or optics tube) need to be. Do we need to go to 3:1 bandwidth? Clem: make some trade-offs for cost vs. having independent telescopes for bands. Get cost for how much it does or does not help to put more bands for multichroic detectors in.
- ACTION ITEM: Study the question of cost versus broadband.
- Jeff's slide on SPT panel gaps: covering the gaps with gap-filler strips, still get panel gap scattering at the -20 dBi level (and lower). Jeff took this beam map and convolved with some image of the sky and found that the galaxy could come in as 10\% of the signal at $\ell=100$ for $r=0.01$.
- Steve's slide on monolithic reflectors vs surface error including deformations. Shows wind deformation calculations with some estimate of how stiff the geometry makes it. DVA1 is a prototype for SKA in Canada, and ATA is a single piece reflector. Dashed are temperature gradients, thick solid are gravity, and thin lines are wind deformations. These don't specify the $\ell$ range. The CCAT segments were 0.5~m. The overall point is that the stability of these large panels is not too bad -- could imagine building a 6m monolithic mirror.
- John C is the expense for the CFRP in the mold, so a first-time cost? Steve -- there is a lot of hand work with each large telescope because distorts after you pop it off the mold the first time; could easily rattle through 5-10M\$ in CFRP development quickly. Metal might be a better option. Could go to COSPAL for 4-5m mirrors with 100 micron rms; could then finish off better, perhaps.
- Temperature gradients are 5-6 K for ALMA; would need to do something to make it more stable like maybe pumping water around, says Steve.
- ACTION ITEM: Study trade-offs and costs for monolithic designs: materials, quantity discounts, surface finishes, stability.
- Ki Won's slide: shows improvement of forebaffle on a beam.
- Have to worry about changing emission from changing temperature of forebaffle, and keeping it dry.
- Tom Essinger-Hileman's CLASS slide with reflective forebaffles. Made it reflective to improve loading; Steve mentions this is rather like CBI baffles which accepted a lot of the beam and so had to be reflective. Clem says for BICEP it was a design choice to accept the extra loading for the benefit.
- Clem says for QUAD their ground screen was possibly detrimental. Lyman says for ACT that is not the case; among other things there is a mountain nearby. Clem -- horizon at SP is featureless enough that you might not need one. Steve says varying snow buildup can be problematic at, say South Pole, with a fixed ground screen.
- Clem wonders how well you can actually remove ground pickup from known sidelobes; has anyone done it like that? Akito points out that if ground pick up varies that is problematic.
- ACTION ITEM: Look at the trade-offs with existing ground screens. Have fixed ground screen in Atacama with ACT, and without in Atacama with PB, and without at Pole with SPT -- Nils suggests we can compare these.
- Issue of RFI and magnetic pick-up ... need to study.
- ACTION ITEM: feed this question back to the science definition: Is it necessary to make a full unbiased (unfiltered) map of the sky (BICEP can get rid of magnetic pickup with their ground synch removal, eg, but lose sky modes)?
- Is it possible to share a common mount among different telescopes (ie, like the full QUIET design to hold several small ones, but could be one big one), asks Steve? Can be studied as things move along.
- Osamu Tajima's slide -- rotating telescope.
- Can rotate in azimuth continuously and/or can rotate about boresight (even in a step-wise manner).
- Sky rotation is not perfect, so for example different sides of an array see Sun differently; shows leakage improvement for QUIET when include boresight rotation.
- Shows successful He hoses through a rotating joint for the small telescope example of Groundbird.
- Advantage of continuous azimuth rotation is you don't waste time turning around, and you keep the instrument stable cf not having to undergo deacceleration; also probably improves loss of modes due to baseline filtering.
- ACTION ITEM: come up with machinery to do simulations based on scan strategy and telescope side lobe estimates (including stability of the sidelobes -- to temperature, wind etc, and stability of source when it is the ground rather than the galaxy, eg.) Feed back form science definition question on filtering from action item above (ie because we will have to filter at some level to correct.)
- PRINCIPLE: Need to decide on what things you design in to the instrument and what things you decide to correct later. General idea is to look at strawman instrument designs that will require minimal corrections and filtering, and their costs, and then other designs with more corrections/filtering and their costs, etc.
- Steve -- obviously very important to leave no mode behind for the reionization.
- Akito -- have to worry that as you push the sensitivity down you will have to filter more and more His simple picture is that everything is like $1/f$ noise so you obviously hit that eventually as you push the white noise equivalent lower and lower. Clem thinks this is not true -- thinks you could scale BK and get down another magnitude. Suzanne notes that historically as you get to new levels of sensitivity, new systematics emerge to be treated.