SLAC-2017:Sensors and Readout Parallel Session

From CMB-S4 wiki
Jump to: navigation, search

Back to SLAC-2017 main page

Optical Coupling, Sensors

- Presentation pushed to Wednesday (we'll try to cover this if we have time. My apologies if we can't cover it on Tuesday.)


Notes from session

System Engineering:

  • Understand instrument requirements. What measurement requirements do we need to specify?
  • Ultimately 1/f comes from low-ell noise requirement, but of course there is interplay with modulators.
  • N_\ell per band. Adrian: there is a superset of bands on the S4 website.
  • Band location:
    • We can define max and min band edge locations?
    • How well do we need to know the bands?
  • Readout:
    • Crosstalk
    • Dynamic range of optical loading
  • Testing requirements:
    • Jamie: beam measurements are critical for systematic control. For example, would we want a dual TES?
  • Some other requirements we will drive:
    • time constant,
    • focal plane temperature stability - From the bicep group, found it useful to temperature-control the focal plane. Use NTD detectors to read the temperature. Written up in first round Suzie paper, and a paragraph or 2 in the BICEP paper. Can achieve focal plane nanokelvin sensitivity - the question is how much power you want to deposit. The question is how much power you get through
  • Requirement on detector-detector angle - probably we don't care.
  • Interfaces:
    • cryogenic thermal budget of readout

Kam: Simon's Observatory plans

  • Goal to develop universal focal plane module that can be used for multiple technologies
  • Full fab at NIST and Berkeley and possibly commercial
  • Pursuing for DfMux and umux. Want to make modules universal
  • Will have CDR

Clarence: SPT-3G and how to get to stage 4

  • Deployed 10 wafers
  • Production throughput is high: 5 arrays every 1-2 weeks
  • Fab time comparable to testing time
  • Film properties vary more than detector architecture: dielectric thickness, Tc
  • Need way to evaluate stability of fab without full array cryo testing
  • Discussion on testing all detectors:
    • enough intra-batch stability?
    • Use stage 3 receivers to have high throughput?
    • Dark vs light test?
    • Probably not end up testing every wafer? Now until CMB-S4: build enough program to give us enough confidence to do this.
  • Importance of testing as part of the R&D, need to be talking about this as a community
  • Jamie: what film properties in particular? small veriations in \epsilon & t.
  • Tc of Titanium for thermistor wonders around.
  • Tested at least 30 6" wafers
  • Kam: List of verification vs. validation, and what needs to be validated on all detectors vs what needs to be validated on representative detectors
  • Clarence: not just measuring a central value, but measure the scatter.

Jay: NIST update

  • TES:
  • mkids: mkids for BLAST/sub-mm and mm
  • umux developing, mustang2, beginning on integrated fab
  • Si platelet horns
  • packaging: Considering 2D backside (e.g Bicep3) or integrated fab/mkids
  • Repeatability has gotten very good - built many test structures to develop processes
  • Brad: assembly takes as long as fab: 1-2 wks per batch
  • Zeesh: lessons learned from recent fab? Repeatability & uniformity have been good.
  • Suzanne: for ACT, NIST gave them 4 wafers, and they could have fielded 4.
  • Zeesh: how did the uniformity get better? Jason: did lots of dedicated process development to deal with uniformity
  • Taylor: how did you get to 4 good wafers on first try? Suzanne: first 4 that were tested all the way through.

Toki: Hybrid fabrication

  • Trying out commercial fabrication with HYPRES based on PB2 detector
  • AlMn done at LBNL/UCB
  • HYPRES can handle the throughput - machines run 24/7
  • Low cost
  • Kent: what about StarCryo? Toki: wafer dirty (contact mask only), uniformity not as good.
  • Brad: Will Hypress do the AlMg fab? Yes, but takes investment of time and money.
  • What other steps not done at Hypress? XeF2 release - they didn't have machine.
  • Jeff: when you say they all work, what do you mean? Toki: next step - hit exact parameters using test chips
  • Peter: does Hypress have facilities to handle throughput? Toki: yes.


  • Using the ABS receiver to test 128 Al LEKIDS
  • Single layer Al film on 160micron fabricated at JPL
  • On-sky testing locally (not in Chile)

Multi-chroic MKID development:

  • Coupling the OMT to an MKID
  • MKID is CPW lambda/4 resonator

Other MKID developments:

  • AlMn LEKID have high Q
  • AlMn allows going to lower frequencies (30 GHz), and higher signal to noise
  • Larger LEKID arrays fabricated at JPL and BNL.
  • Magnetic field create vortices in the MKID. Probe tones move those around and degrade the Q.
  • cross-polar response - doesn't get to zero.
  • Fab source for the single-layer LEKID was doen at JPL
  • Fab source for multi-chroic MKID is SLAC
  • AlMn can be used for MKIDs, which is good for lower frequencies
  • Fab from JPL, Brookhaven, and SLAC
  • Jamie: status with 1/f? <0.1 Hz. Now dealing with systematic effects in the measurement

Jeff: Horn-coupled architecture

  • Push the horns as far as possible
  • Quad-ridge waveguide reduces footprint
  • Chao-Lin: what is benefit? Jeff: it's that the planar sections of the OMT

Jay: umux

  • umux repeatability improved from fab processes
  • flux ramp line used to linearize SQUIDs
  • Noise subdominant in Mustang2
  • TES bias is done in groups with interface chips - same as TDM
  • What is status of 1/f noise? Jay: this data is not focusing on 1/f noise. Kent: it's below 1 Hz, but we haven't looked closer.
  • Rely on TES properties to be uniform? Need less uniformity than with FDM. Able to bias anywhere in the transition. Hasn't been a practical constraint.
  • Isn't it the bandwidth available? Different sets for different colors.

Jeff: Integrated fab on the wafer

  • 6 connections per wafer
  • Need to remove nitride layer to fabricate umux components
  • Minimize component size to maximize packing
  • Are you worried about SQUIDs being in the middle of the focal plane? Kent: these are individual elements - you do carry about pickup, but these are really small gradiometric baselines

Akito: on-wafer DfMux

  • Simple LC resonator needed to be added to detector to make it on wafer
  • Size of LC resonator needs to decrease in size for integration. Can go to higher frequency to reduce size
  • IDC capacitor currently. Can go smaller with parallel plate capacitor (SRON did)
  • Kent: noise performance at 20 MHz? Akito: don't have data
  • What about capacitors? interdigitated is what they do now, but parallel plate reduces size. Hydrogenated amorphous silicon as the dielectric at SRON was low.
  • Kent: relative focal plane area usage: blue square is L, blue square is C.
  • Suzanne: for a parallel plate capacitor, does that increase fab complexity? Toki: only one more layer.

Sarah: SLAC RF electronics

  • Multi-purpose RF readout
  • Designed to have linearity and noise to handle 4000 channels in 4-8 GHz
  • 2.5 Gs/s direct synthesis.
  • Reduces power going to HEMT by order of magnitude.
  • Kent: allows you to have flux-ramp mod use the full bandwidth of the resonator.
  • 4-8 GHz has 8 blocks of 500 MHz. You could of course directly digitize for 0.5-1.0 GHz.

Action items/Next steps

  • Agree on list of input and output for each sub-group
  • Make table, then start filling in input and output for given technology. Input and output maybe in parametric form (for example beam FWHM vs pixel size, mapping speed vs receiver and sky temp)
  • Start discussing throughput. Detector fabrication, assembly, test. Feed back cycle time