John,
I haven't seen the existing data, so I can only speak in general. The TES transfer
function should look roughly like a single pole with a time constant of order ~1 ms (f3dB
~150 Hz, 1-pole), probably capping out at maybe ~20 ms for a nearly-normal detector.
SPIDER measurements have not been very extensive, but we have seen evidence for subtle
structure in SPIDER's electrical transfer function at somewhat lower frequencies.
All of this should have only a tiny effect in the science band. We're totally
dominated by the MCE (f3dB~60 Hz, 4-pole) and GCP (f3dB~10 Hz, many poles) transfer
functions, which we can model analytically.
Higher-bandwidth data has the advantage of letting you measure the intrinsic TES transfer
function more accurately at high frequencies for model-fitting and such. The dream data
set would be optical response at a variety of biases and a range of sky-like loadings,
which is obviously hard to acquire and analyze. For science analysis, however, I agree
that you probably don't care about all of that and that GCP data should be fine. We
should be able to measure the small deviation from the GCP/MCE transfer function
reasonably well with GCP data. I don't have a good sense of how much the TES
transfer function will vary during the observing runs, but I would hope it is small?
Justus has put a lot more thought into this and I haven't run numbers for the
magnitude of the pointing offset from the TES transfer function, but at first blush I
agree that we should be able to work with GCP data.
- Jeff
On Dec 18, 2012, at 6:50 AM, John Kovac <jmkovac(a)cfa.harvard.edu> wrote:
Hi Justus and Jeff,
Justus, this is the last day to consider the questions in my email below. What I am
hoping to hear from you is a well-reasoned confirmation that the transfer function data we
have in the can is all we are likely to ever need to support BICEP2's B-mode science
analysis.
From your previous reply, it sounds like we have complete GCP-acquired data with the
2011/12 biases, and some MCE-acquired data for the original 2010 biases, presumably both
under sky loading. I don't know of a posting that shows the limitations of GCP vs
MCE data for constraining transfer functions in our science band. I'm operating
under the assumption that the GCP-acquired data is sufficient for our science analysis,
which seems likely to me, but I really don't have the information to be sure so
I'd like to hear from you.
Jeff, can you help shed any light here?
If there is a need to retake sky-loading transfer function data with the MCE on the DSL
balcony it will be a lot of effort for folks at Pole in the next several days, but in
principle there is still time if we decide today. Let's try to get a firm decision
at the telecon in 3 hours.
John
On 12/13/12 4:00 PM, John Kovac wrote:
Hi Justus,
I'm afraid it is really too late now to take additional data on the mount--the team
at Pole is taking BICEP2 off the mount within the next 2 hours. We can beg them to point
BICEP2 at the sky off the DSL balcony, but if so we have to really understand why
we're doing this.
Why do you want to see the raw detector transfer functions, without the GCP and MCE
filters?
What is the value of the MCE-acquired data to our science analysis, compared to the GCP
data? All that matters for our ultimate analysis is whether we understand the transfer
function well enough in the science band, right? Does your old MCE-acquired data show
evidence that some detectors have raw transfer functions we may need to correct for even
at < 5 Hz?
Or is the GCP-acquired data taken with 2012 biases/tunings on the sky good enough to
exclude the possibility of A-B transfer function mismatch in the science band for our RGL
detectors at the level that would introduce significant T->P contamination? What spec
on this have you been targeting? Roughly speaking I would expect it to be similar to
relgain mismatch--that was the argument we made in Takahashi et al. I'd like to see
what analysis you've done--have you got GCP-acquired measured transfer functions for
all the RGL's?
John
On 12/12/12 12:23 PM, Brevik, Justus A. wrote:
Hey John,
I've been looking through the transfer function data that I acquired through GCP,
and the data quality looks very good and I believe I have a complete set of all the bias
and tuning settings we used. Well, at least what we used up until Feb 2012. I'm not
sure if we have used any new tuning (possibly) or bias (unlikely) setting since then. If
we have, then perhaps we should run a schedule for that data. I should hope that
different tunings (and their effect on the flux feedback servo) should have a minimal
impact on the transfer functions.
In addition to looking through the GCP acquired data, I have been searching for transfer
function data acquired directly through the MCE. I had planned on taking this data last
Feb but it doesn't appear that I did. Some of this data exists from the previous
year, but not at the new bias / tuning settings. It would be useful to have this data in
the can in order to see the raw detector transfer functions, without the additional
filters added by the MCE and GCP.
If there is still time to take this data, I think it would be prudent to spend some time
on it. I can write a note on how to take this data for Jon (or whomever else), as well as
get on the phone or gchat to work with him real-time, and do spot checks of whatever data
I can get my hands on. Do you think there is time and motivation to do this? Sorry this
data doesn't already exist in a complete form.
Justus
--
___________________________________________________________________
John Kovac jmkovac(a)cfa.harvard.edu
Assistant Professor, Astronomy and Physics, Harvard University
160 Concord Ave rm 310, Cambridge MA 02138, 617-496-0611