LBNL-2016: Instrumentation I

From CMB-S4 wiki
Jump to navigationJump to search

Session Organizers:

Overall: Adrian Lee


Detector Arrays and Readout: Jamie Bock Kent Irwin Hannes Hubmayr (can't make it), Jay Austerman will be there from NIST Clarence Chang Adrian Brad Johnson Erik Shirokoff

Optics: Nils Halverson Mike Niemack Shaul Hanany (can't make it) Steve Padin

Polarization Modulators: Akito Kusaka Toby Marriage Shaul can't make it... Chao-lin Kuo

Summary of optics/instrumentation discussion from March 8: File:Instrumentation1 summary 20160308.pdf


File:S4 telescope optics questions 20160307.pdf

1) For small, medium, and large apertures, what throughput can we get from each family of design e.g. reflector/refractor, cross-dragone

Some throughput comparisons for medium to large (2.5 - 10 meter) apertures are summarized in these slides: File:Niemack Optics 20160507r.pdf, more details in [Mike's paper]

File:SPT3G Optics Overview S4 Workshop 20160307.pdf


1) There is a need to have a simulation loop that incorporates realistic instrument characteristics. - We can use stage 3 data, e.g. deep beam/sidelobe measurements. - We can use simulated properties, e.g. far sidelobes.

An important question is the timeline of the iterative process of design. For example, once we go through a design loop, should we make a prototype? or prototypes?

Mike Jones: KSA: requirement->analysis->prototype->iterate with malleable design

2) We are getting relevant stage 3 data, maps at < 5 microK*am.

We are getting stage 2 and 3 data on several types of instrument. Medium or large apertures, with/without rapid modulation.

3) Need some strawmen to simulate.

what beam sizes are needed at what frequency since it drives the design -> need input from forecasts.


assignment was to discuss optics, but optics design needs input from the instrument.

a question: how many telescopes do you want practically just 1 or tens?

2) For each science goal and fixed resources, is there an optimum in the aperture vs. detector number tradeoff?


3) What is the verification path and timeline for polarization modulators?

a) Stepped vs. continuous?

b) What systematic errors do they mitigate? Do they introduce new systematic errors?



4) Assuming a heterogeneous array of small, medium, and large telescopes, what are the specifications required for each?

What are the largest systematic errors and how do we mitigate them? Near and far sidelobe performance (e.g. panel gaps File:MonolithicMirrorsPadin0316.pdf, ground shields,...),

What tests are required to measure the low-ell limit of medium and large telescopes?

DETECTORS AND READOUT File:DetectorsAndReadoutV4.pdf

5) What are the detector specifications required for CMB-S4?

Answers will be driven by science, foregrounds, observing strategies and telescope design.

How much do detector specs differ for different telescope platforms (aperture, optical configuration)?

What do we need to specify bandwidth, beam symmetry, etc.?

6) Timeline

Kathy Turner has announced a “small funding wedge” in 2018. It seems likely that significant funding won’t happen before 2020 from the DOE.

It makes sense to develop technology options at least until project funds turn on, and maybe longer.

What is the appropriate timescale for the development of technology options?

7) Cost What are our cost targets for production of detector parts?

O($100 / pixel) is too much

O($10 / pixel) would be fantastic

O($1 / pixel) is way past the point of diminishing returns to CMB-S4

Somewhere between $10 and $40 per pixel is probably the target. Akito assumed $40/pixel.

Is this the right way to think about it? In the DOE budget there will likely be FTEs to support fabrication teams, rather than funding allocated per pixel. This also doesn’t include infrastructure cost.

Mass testing/characterization of detectors is also a critical cost driver.

What is the appropriate balance of investment early on to reduce production costs, system complexity, and test/characterization cost?

8) Downselects

Does CMB-S4 use the same sensors, band-definition filters, beam-forming elements, cryogenic multiplexers, and readout electronics on all arrays?

Political driver from DOE culture: downselect to one technology option.

Political driver from an energetic and diverse community: use many technology options.

Can we find the right balance of Science, Risk, Cost, Politics?

Totally homogenous arrays?

Different technology choices by frequency?

Different by site?

Different by aperture?

9) Polarimeter technology options

Sensors: TES (mature), HEMT (mature), MKID

Array architecture / beam forming. Lenselets, Feedhorns, Phased planar antenna

Bandpass definition: Multichroic? How strong is the driver for more than two frequencies?

Advantages / disadvantages Sensitivity – photon noise dominated + NET (efficiency) Fabrication complexity, yield, uniformity Optical efficiency, crosstalk

How can the community coordinate? What will the decision process look like?

10) Cryogenic multiplexer options

Mature MHz multiplexers: TDM, FDM

Emerging options: MKID, Microwave SQUID, Parametric amplifier + GHz resonator, KPUP + MHz FDM

How valuable is it to integrate multiplexing elements on the detector wafer? It brings significant reduction in integration complexity. Is it necessary, or advantageous?

Room-temperature electronics: Mature Canadian TDM + FDM, Immature American ROACH, GPU, LCLS, etc.

Advantages / disadvantages: Detector-limited noise (vs. preamp, DAC, digitizer noise), Cost. Reliability, Scalability.

How can the community coordinate? What will the decision process look like?

Summary of detectors and readout

We discussed what we know today of the pathway for detectors to the requirements of CMB-s4, including specifications, timeline, cost, plans for downselects, and technology options, especially in readout.

The detectors consist of both the polarimeters and the cryogenic multiplexing circuits.


1) Detector technology should be advanced for CMB-s4 at least until significant ramp-up in CMB-s4 project funds (possibly 2020?), and potentially longer. These advances can reduce cost, risk, and complexity of the CMB-s4 arrays. It was suggested that the detector development community should get together periodically to track progress.

2) The production cost of the stage-3 detector arrays must be reduced to scale to CMB-s4, but attention should be paid to optimizing the overall cost of the detector arrays including both the cost to build up the production capability, the support costs of the fabrication teams, and the marginal costs of producing the detector array components. This will be expensive.

3) Testing and characterizing large numbers of detectors will also be a cost driver that should be watched.

4) Decisions will need to be made as to how the CMB-s4 detector technology will be downselected. How homogeneous should the CMB-s4 detector arrays be? Should they all use the same sensors, optical coupling elements, and readout, or might there be different optimal solutions at different frequencies, or different apertures, or different sites?

5) We talked particularly about readout technology, and considered whether existing MHz FDM and TDM circuits can be scaled to the needs of CMB-s4, or whether advanced versions of these technologies, or emerging readout technologies such as MKIDs or microwave SQUIDs, will be necessary. We discussed whether it is necessary to integrate the multiplexing elements onto the detector wafers, or whether interconnects can be sufficiently automated to scale existing approaches.

There was controversy on this point, with some feeling that the complexity of existing readout circuits, including particularly connectorization, will make them difficult to scale, and others feeling that this should be doable with appropriate automation. We concluded that an analysis of the cryogenic load and complexity of existing approaches should be studied, and readout approaches with higher multiplexing factors and integration in the detector wafer should be developed.


Clarence Chang, Zeesh Ahmed, Adrian Lee, and Jeff McMahon will lead a group to what we can define about the science case for pixels with larger bandwidth coverage and more multichroic channels. Getting an early idea of the drivers for bandwidth and multichroic channels can help focus detector R&D.

Break-out sessions:

Wiki navigation

Return to main workshop page