Back to Argonne 2018 main page
Parallel session P2-1: Readout (chair: A. Bender)
File:S4 readout intro.pdf (5 min): Amy Bender
Connections to forecasting (5 min): Colin Bischoff
File:20180305 S4 fmux2.pdf - Technology Comparison: Effects on statistical power and systematic effects (10 min): Adam Anderson & Shawn Henderson
File:Readout Cost Schedule.pdf - Technology Comparison: Effects on cost and construction schedule (10 min): Hannes Hubmayr, Darcy Barron, Tyler Natoli
Discussion & Formulation of Reference Design (30 min)
Notes from session (Tuesday, March 6, 09:45-10:45)
Homework from the session
- Cost out the three options
- Evaluate the risks of the three options
Slides *Colin on Forcasting *Much coming out of forcasting for r, and looking to merge forecasting for r and other scienc ⁃ Have not previously had reference design, which has led to broader choices and current results are disconnected from actually instrument ⁃ large number of actual instrument characteristics could be causing the effects being modeled ⁃ challenges for next round of forecasting ⁃ hope to get more specific on certain effect with reference design better definged ⁃ trying to match work going forward with topics in TD lists as much as possible ⁃ have not done any time ordered simulations yet. ⁃ need to think about how to get effects into a map based form that is easy to understand. Haven’t done this yet ⁃ For CDT report there allowable thresholds on r, but this needs to be expanded beyond r which is not the ultimate answer, also need to look across multiple frequencies ⁃ Statistics and Ssytematics impacts of different technologies (Adam A) ⁃ fmux current status: 68x with SPT-3G and 40x on PB2 ⁃ using same warm electronics can be extended to 128X and similar cold electronics ⁃ can impact cost, but not directly sensitivity and systematics ⁃ umux ⁃ should have very high multiplexing factors - baselined for SO ⁃ not used yet in CMB, 64X in mustang, and similar to MKID readouts ⁃ nice results in Dober et.al. 2017 showing low noise, cross talk, and 1/f knee ⁃ Gustavo - warm electronics have been tested on sky for 10/20K and have gather for optical applications. Adam, but the MKID/UMUX is not identical ⁃ Side discussion on relative value of DC/AC bias. Both have been used in demonstrated experiments, no clear that it has a dominant impact on the instrument ⁃ Readout cost and Construction (Tyler N) ⁃ FDM is know cost structure, can scale up from 3G / PB ⁃ need to be mindful of labor costs ⁃ UMUX ⁃ no current stage 3 using it. Has been costed for SO, but not fully built out ⁃ pros are highly scalable and less tedious to put all the components together ⁃ For costing, it will be good to look at contingency levels for different technologies. Also need to specify using current cost estimates versus possible reduced costs in the future due to equipment advances. ⁃ General agreement that on the full scale of the project, readout cost is not a main cost driver
⁃ Discussion ⁃ What are the requirements for a multiplexing scheme to be chosen for the reference design? On-sky, in CMB bands, across all CMB bands? ⁃ TDM is only solution shown on sky, over all bands ⁃ FDM is shown on sky over many bands ⁃ UMUX is on-sky, but not in CMB bands ⁃ Suggestion and general agreement that readout should not be dominant noise source (I.e. detector or sky limited) ⁃ this is not as simple as it appears, and can be hard to make apples to apples comparison regarding bands, fridge temps, etc. ⁃ Amy suggests that it would be possible to consider using different technologies over different bands (later voted on not to do this) ⁃ Discussion of multiplexing factor and cost ⁃ not clear on what mux factor is required, which would drive cost ⁃ agreed that mux factor is not the main driver of total instrument cost ⁃ Gustavo cost for Fermi electronics is $7 per channel, but biasing is not included ⁃ Brad expressed concern is that this is the highest risk area. Least risk is TDM, but no one is pushing this ⁃ Zeesh, could list TDM for the reference design, but then we have two options with risk to be reduced ⁃ concern that listing TDM if we are not seriously invested in pursuing this could lead to trouble in the future ⁃ issue with if anyone in CMB community is interested in pushing R&D on this ⁃ question on if TDM could be built to CMB-S4 scale in time frame given. General sense that this could be done ⁃ Kent - I think this is over estimating enthusiasm. Only new one is UMUX. But people still work on FDM ⁃ Brenna, anything demonstrate on sky to the level needed. Crowd - TDM ⁃ Discussion on cost of readout going down over time, especially as technology advances ⁃ could be a bigger drop in telecom band technlogy ⁃ again not the overall driver of cost ⁃ focus should be where we stand now for reference design - should be shovel ready ⁃ Toki - whatever we pick needs to be shovel ready - not sure we can claim UMUX at this point ⁃ Discussion on making a selection for reference design ⁃ Tyler - is it a big switch to baseline one in reference and then switch to something else in final design ⁃ Amy - that might be a good way to approach it. Have a reference design and a strong option, Both will still be pursued ⁃ Adrian - reference design will be an evolving document ⁃ Clarence - strong issue is costing. If we baseline something that is much cheaper then the real option, that could get us into trouble. Would be good to check on this for the future ⁃ Toki - wasn’t this done on SO and seen to have similar cost. Agreement from Zeesh, mux factor of order to 500 and fdm of order 64 in SO cost ⁃ Brad - will need much larger contingency for readout ⁃ Darcy - also should look at reason for mux, cost, power, and dissipations (there were 3)? ⁃ Brenna - at some point you will need to demonstrate what that these systems work before CD-3. Should leave TDM as an alternative if it is the only option that has been fully demonstrated ⁃ Zeesh - so the feedback would be to pick one that isn’t fully proven, but then use TDM as a fallback option. ⁃ Brenna - yes, can list one of the items that is really exciting and we want to move to, but have TDM as a REAL option. Reference design needs to be a fully self consistent design ⁃ Toki - be careful mixing up readout options. Also have to think about testing facilities ⁃ Amy - make a motion that we want a single technology as reference across all frequency bands with others at options ⁃ Amy L seconded ⁃ 22 in favor ⁃ ? opposed ⁃ 1 abstention ⁃ Kent is there an argument that different frequencies need different technologies (packing factor) ⁃ Adrian, could go to smaller FDM resonator size, but is possible ⁃ Brad - yeah there would be work needed ⁃ Clarence - at lower frequency bands moving to other options could be a different path ⁃ Amy - what pieces of information are needed to make this decisoin ⁃ Brenna - would be good to have a few people develop the options for different cases and present this for later ⁃ Adrian -for the poeple in the trenches there seems to be good understanding on the basics and have been discussed ⁃ Suzanne - would people be willing to take a non-binding vote to gauge the temperature of the support to see if we are truly split. Only for reference design, for all bands ⁃ almost perfectly splits among all four (TDM, FDM, UMUX, no vote)
Notes from Wednesday discussion
Homework from Tuesday (during Wed discussion):
- Costing for all three technologies (to make sure they are ~similar).
- Padin: Learn what information he needs as input, make sure three techs are costed similarly. Zeesh can provide inputs for TDM.
- Some assessment about scalability for all three systems.
- e.g., number of wires, mUX factor, number of SQUIDs /LNAs, cross talk, linearity, stability, loading on mK and warmer stages, number of flanges
- Performance specs for different technologies (tMUX, fMUX, uMUX)
- Kent rec: Readout noise, photon noise, detector noise referred to input, MUX, and on-sky.
- Amy (fDM), Zeesh (muMUX), responsible for assembling specs. Need tMUX person. Get Padin involved. Cite references or private communications / plots.
- Get an external reviewer to see if anything is missing.
Suzanne notes it would be nice to have demonstration plots in Ref document in Nov.