From CMB-S4 wiki
Revision as of 12:16, 21 October 2016 by Galtsev (talk | contribs) (Created page with "== '''Overview:''' == We have time to think. The goal we came up for this whitepaper was to summarize. Goal now is to start planning R&D efforts. Start prioritizing R&D for...")
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigationJump to search


We have time to think. The goal we came up for this whitepaper was to summarize. Goal now is to start planning R&D efforts. Start prioritizing R&D for the next 2 years. Common theme: lost of people have done really amazing work, with lots of progress in the past 5 years. This is the opportunity to select the best technology to solve a given problem.

Cold optics

Notes on content during presentation that isn’t in the slides:

People have always wanted modulators in polarization modulators. Some interesting polarization modulator ideas are coming on-line right now.

Don’t yet have requirements, but have targets for what we want in our toolbox. Lots of different instrument configurations could be used, but still you can imagine optical elements from 30 cm to 1 m.

Another comment on window development: Investigate windows for a tiled design.


Steve: what is your sense on performance vs diameter. Jeff: when things get thicker, they get more lossy. This comes to a system-level optimization.

Carlstrom: I don’t see problems shown in the whitepapers. Jeff: when people said this was ready, there was push-back. Suzanne: there are some hard truths in the mux section…

John Rhul: once you’ve adopted a high-index lens, is the addition of a waveplate a throughput-limiter? Jeff: right now it is not, either for sapphire or silicon.


Notes on content during presentation that isn’t in the slides:

Spectral resolution: Holzapfel: on splitting atmospheric bands into two spectral bands, why do you care if the bands overlap? Kovac: forecasting assumes overlap anyway. Adrian: maybe not so important, but want to avoid lines also.

Slide 5: Note that reduced mapping speed at “foreground monitor” channels may be perfectly acceptable, since foregrounds are brighter there.

Slide 5: integrating readout multiplexing on wafer would reduce interconnect complexity.

Slide 6: the directly-coupled LEKID addresses interconnect issue by having readout on same wafer.

Slide 9: Toki: Should we test all the wafers? People say yes. Toki wonders if this is true.

Slide 10: Toki: for notional detector array designs, just need to get some logistical constraints


Ruhl: in SPT-3G, using all bond pad space around perimeter. 1500 detectors on a 6-inch wafer. Is the perimeter constraint actually strong that it limits multi-chroic.

Kovac: cost trade off of getting as much as possible out of focal plane area, or making focal plane area cheap. How to view this trade-off? Jeff: how much does a telescope cost? Pryke: just driven by the detectors. Keck receivers are way less than 400k (outside detectors). Kovac: cost not just in terms of money, but also complexity. Zeesh: this sort of costing has not occurred yet. Adrian: if your silicon area is driving the cost, adding more complexity to silicon could be good. If silicon area is driving, it might be good to add complexity to that silicon.

Stark: how much could detectors be improved by improvements in efficiency or loss. Is there a need for R&D? Toki: no, mass production is really the key R&D. Suzanne: goes to John

Sensors / Readout

Notes on content during presentation that isn’t in the slides:

Slide 5: Zeesh: prime scalable technology

Slide 6: Zeesh: mux factor of 500 is probably the sort of number that makes sense for CMB-S4, but varies by technology

Slide 6: Zeesh: what does a demo mean? Bill: do we have a set of requirements for noise floor, 1/f? Zeesh: this is on the “to do” list.

Slide 6: Zeesh: muxing with KIDs easier than with TES, but needs on-sky demo to validate performance/noise.

Slide 10: scale up a warm cross-platform electronics


Keith: good points about KIDs, but one question is, is there an early sense of whether getting very high yields will be easier with KIDs than TESs? Brad: we’ve achieved 90% yield with KIDs.

Hannes: Comparing yield. TES arrays are much more complex. If direct-absorption single-band is on the table, KIDs are a great solution. Multi-layer can be done, it just hasn’t been shown. Is there a significant difference in quantum efficiency? Irwin: two different questions – how much goes into quasi-particles, and how much gets there to begin with

Hannes: what about lower freq? Brad: it’s possible. Kovac: but you don’t need as many detectors. Brad: also it would have to be

Clem: didn’t talk about crosstalk! Zeesh: we should define a requirement. In the generation of experiments right now, the crosstalk number we have now is 0.1%(umux), 0.3%(TDM),1%(FDM).

What’s the fundamental limit on the uMux? Zeesh: practically limited only by packing and crosstalk. Irwin: basically the same as KIDs.

Kovac: KID on-sky demo. What does this entail? Make a map that shows the noise integrates down. Devlin: KID and uMux receivers are fairly simple, so you can swap in a

Telescope designs

Notes on content during presentation that isn’t in the slides:

Slide 2: how low in ell the large-aperture telescopes can go is one of the things we hope to show in the next 2 years.

Slide 5: S4 needs to have a much more systematic error control


Debroulie: what is limit on low-ell for large apertures? Niemack: we don’t know. Hopefully polarization modulators will help.

Jay: building large telescopes – how long does it take? When do we have to have a large telescope design fixed? Devlin: ACT/SPT were both 3 years construction.

Clem: everything here made sense. Systematics at the level we need will be fantastically hard. It will be hard to do anything other than “conservative” based on simulations. Do we need a measured prototype antenna? Clem can’t see how we could decide to build a lot of them at once. Niemack: well, new telescope designs are a lot more powerful. So we’d like them. Do we want to make

Adrian: hybrid design with large and small apertures satisfies dynamic range difficulties, and doesn’t rely on large apertures for low ell. You can imagine scaling the small apertures a little bigger. Do lensing, but keep a lot of the systematic error controls the same.

Stark: if you’re able to make an antenna with such a large field of view, you can afford to make the telescope more complicated and expensive. Then you can have an actively controlled. For example, correcting for gravitational warping could be active.

Final discussion for the entire instrument white paper groups:

Nils: do we need “systems engineering” as a separate topic? It’s going to take a lot of mundane control looking of reality of the entire system. For example: simulations that take into account achieved sensitivity on bicep/keck. We could use our knowledge from fielding actual instruments. Taking into account non-idealities as an up-front design problem could help.

Clem: let’s all out our noise spectra. What’s limiting the ability? If person-power is all it is, let’s get that out there. Lawrence: low-ell in polarization are almost impossible. Everything is harder at low ell.

Jeff: a systems group might help with putting the technologies together.

Suzanne: is ell=30 all that we want to hit? Clem: could characterize it as “should S4 grow another effort to go to ell=5?”

Ruhl: at some point our thinking about configurations will be constrained by cost. How close are we to that? Kam: depends on development. Adrian: how to prioritize R&D now? For example, if you know it’s really important to focus on

Devlin: there is a theme of increasing complexity, but maybe when we look at a system level we’d see that this isn’t good.

Suzanne: should we agree tomorrow for getting a systems group together.

Zeesh: from detectors and readout, we want to identify execution of R&D really soon.