Hi Danny,

It is an internal document, I am going to ask if it is ok for me to forward it 
to you, probably it is fine but I will double check.

I am trying to think about this one a bit more… I am not sure I fully 
understand some of the reasoning behind it.

Cheers,
Eloy.

> On 13 Oct 2015, at 19:50, danny jacobs <[email protected]> wrote:
> 
> Eloy could you send the link to this full paper? I'm having trouble finding 
> it (Stephan has a lot of papers about "calibratibility", the most relevant of 
> which SKA-TEL.LFAA.SE.CAL-AADC-TN-001- seems to be not available online in 
> the form listed on his website.)
> 
> Let me see if I can summarize the arguments.  Spectral leakage from 
> foregrounds can be completely eliminated at the sensitivity of a 10 minute 
> integration by a combination of direction dependent calibration in the 
> primary fov and by the sidelobes being very low outside of the fov. Thus the 
> only physical requirement is on the beam inside of a 10 minute coherent 
> average on an EoR imaging slice of 1MHz.
> 
> Where I'm a little confused is how we can get foreground subtraction to EoR 
> precision at a 10 minute cadence. Suppose there is a spectral ripple in the 
> feed at ~-25dB, below the noise level of a 10 minute integration. We can 
> integrate longer to try to calibrate it out, but technically that violates 
> the spec.
> 
> Theres also a more general critique to be made, that the spec starts by 
> assuming perfection in calibration and foreground removal, something that has 
> yet (as far as I know) to be demonstrated in practice. Presumably there are 
> practical things we can do on the instrument side to make this process 
> easier. 
> 
> Cheers,
> ~Danny
> 
> 
> 
> On Thu, Oct 8, 2015 at 1:16 AM, Eloy de Lera Acedo <[email protected] 
> <mailto:[email protected]>> wrote:
> Hi Danny,
> 
> I think the 60 MHz spacing is not related to delays at all, although I may be 
> wrong… I never heard that argument within SKA. 
> 
> That requirement was derived from the following argument (from a “calibration 
> requirement” document by Stephan Wijnholds):
> 
> ———————————————————
> "
> 
> This requirement is currently formulated as “The stability of the average 
> station band shape post RFI-mitigation shall be within +/- TBD %” [1]. 
> Combined with the requirement on station beam stability, this indicates, that 
> in the envisaged mode of operation of the LFAA element, the station 
> calibration should ensure that the station beam behaves spectrally and 
> spatially according to a specified full-polarization beam model thus allowing 
> sufficiently accurate prediction of the station response during processing of 
> the station signals. To set a meaningful quantitative requirement on the 
> spectral and spatial stability of the station beam, we thus need to see what 
> accuracy is required for array level calibration.
> 
> We expect to be able to exploit our knowledge of station configuration and 
> antenna response to predict the station beam after station calibration. In 
> practice, station calibration will be done with finite accuracy and we will 
> have imperfect knowledge of the antenna response, for example due to 
> manufacturing tolerances. This may result in systematic deviations from the 
> predicted station response. Fortunately, we can correct such errors by 
> direction dependent calibration [2][3] at a cadence of typically 10 minutes. 
> Below, we derive a requirement on the required station beam accuracy for 
> sources inside the FoV. For the impact of sources outside the FoV, we rely on 
> suppression of these sources by the station sidelobes. Combined with the 
> assumed 10-minute update rate, the station beam accuracy requirement 
> naturally leads to requirements on the spectral and spatial beam stability.
> 
> In [4], two distinct approaches are presented that relate the impact of 
> station beam errors to the thermal noise. Although the mathematical bounds 
> used in these two approaches can lead to slightly different proportionality 
> constants (the factor 1/√2 in the equation below), both lead to a similar 
> result stating that the relative errors on the station beam (relative to the 
> main beam peak) should be smaller than
> 
>  <Screen Shot 2015-10-08 at 09.12.54.png>
> 
> where ∆S0/√(B·t) is the noise level of a single station after integration 
> over bandwidth B and time t and Srms is the total RMS flux in the image.
> 
> The instantaneous thermal noise level ∆S0 = 2 kB·Tsys / Aeff follows directly 
> from the sensitivity requirement and Srms can be calculated based on the 
> source statistics from [5] and the size of the FoV of a 35-m station. We will 
> assume that the sensitivity of the full LFAA system will be 100, 760, 1025 
> and 975 m2/K at 50, 100, 160 and 220 MHz respectively and that the data are 
> split in slices of about 1 MHz for EoR tomography [6]. Since direction 
> dependent array calibration can be exploited to correct systematic errors in 
> the beam model at time scales of order 10 minute, we will use t = 600 s. This 
> leads to the following reformulation of requirement SKA1-SYS_REQ-2621:
> 
> Spectral stability. The spectral stability of the station beam bandpass post 
> station calibration and RFI mitigation shall be within 1.3%, 0.4%, 0.6% and 
> 1.1% at 50, 100, 160 and 220 MHz respectively compared to then full 
> polarization parameterized beam model.
> 
> "
> 
> ———————————————————
> 
> 
> Cheers,
> Eloy.
> 
>> On 8 Oct 2015, at 00:48, danny jacobs <[email protected] 
>> <mailto:[email protected]>> wrote:
>> 
>> Maybe there is?  Looking at the requirements in the "SKA phase 1 (level 1) 
>> requirements" I see the following 
>> 
>> "The spectral stability, on a time scale of 600 sec.,of the station beam 
>> bandpass, post station calibration and RFI- mitigation, shall be within 1.3 
>> %, 0.4 %, 0.6 % and 1.1 % at 50 MHz, 100 MHz, 160MHz, and 220 MHz 
>> respectively compared to the full polarization, parameterized beam model.
>> 
>> Translating into the HERA framework:
>> I read this as a suppression of all variation to the  -22dB level at 100MHz. 
>>  Since these numbers are given in ~60MHz bins we could read this as 
>> requiring -22dB at 16ns or longer.
>> 
>> How does this relate to the HERA spec? The HERA specification is -60dB 
>> suppression at reflection time scales of 60ns and longer. (This spec is 
>> rough and will change slightly as we refine it but this is in the  ballpark.)
>> 
>> We can translate the (current) HERA spec to SKA spec language:  Spectral 
>> stability shall be better than 1e-4 % on spectral scales of 17MHz or shorter.
>> 
>> So looked at this way, the HERA spec and the SKA spec are about 4 orders of 
>> magnitude apart.  Recently we've discussed the possibility that the HERA 
>> spec is a little more stringent than we absolutely need, so we'll see where 
>> we land.
>> 
>> 
>> Digging further into the SKA spec:
>> 
>> Looking back even further into the level 0 (science) requirements that are 
>> cited as the origin of the 0.6% spectral stability requirement I find this
>> "SKA1 shall possess relative calibration accuracy of adjacent frequency 
>> intervals of better than 40 dB for Δν/ν=10^-2 to Δν/ν=10^-5"
>> 
>> At 100 MHz this corresponds to a ripple constraint of -40 at 1000ns or 
>> longer.  This one is hard to understand as flowing from a 21cm science case. 
>>  Its also hard to see how that flows into the ~1% error allowable in the 
>> level 1 spec.  
>> 
>> What am I missing here?
>> 
>> 
>> 
>> 
>> 
>> 
>> 
>> On Wed, Oct 7, 2015 at 4:15 PM, Eloy de Lera Acedo <[email protected] 
>> <mailto:[email protected]>> wrote:
>> Hi Miguel,
>> 
>> Yes, understood. I have to say also that SKA does not have any requirement 
>> for spectral ripple in the sense HERA does. I never heard about anyone 
>> thinking about doing avoidance with SKA for example, or at least that never 
>> got translated into requirements. The only concern I have appreciated is 
>> about calibratabilty and absolute A/T across the field of view. 
>> 
>> Eloy.
>> 
>>> On 7 Oct 2015, at 23:28, Miguel Morales <[email protected] 
>>> <mailto:[email protected]>> wrote:
>>> 
>>> Hi Eloy,
>>> 
>>> I’d like to mention that the spectral features we are interested in are at 
>>> the per element not the per station. While per element spectral features do 
>>> average out in frequency, they don’t go away, they’re just moved to the 
>>> station beam (spatial). As we need exquisite control of both, this doesn’t 
>>> really help us. So the golden metrics for me is per antenna response and 
>>> how repeatable it is antenna-to-antenna. 
>>> 
>>> Best,
>>> 
>>> Miguel
>>> 
>>> 
>>> 
>>> 
>>> 
>>>> On Oct 7, 2015, at 8:40 AM, Eloy de Lera Acedo <[email protected] 
>>>> <mailto:[email protected]>> wrote:
>>>> 
>>>> Hi John,
>>>> 
>>>> I am checking with our australian colleagues, it should not be there. I 
>>>> can’t remember now if we ever discussed this. This is also a very small 
>>>> array of course (16 antennas) so it has some features that won’t be 
>>>> present for larger stations.
>>>> 
>>>> Eloy.
>>>> 
>>>>> On 7 Oct 2015, at 16:32, Jonathan Pober <[email protected] 
>>>>> <mailto:[email protected]>> wrote:
>>>>> 
>>>>> Thanks, Eloy.  The feature at 100 MHz in the Y polarization is definitely 
>>>>> the most prominent.  Maybe it's related to FM?  The X polarization also 
>>>>> has smaller features around the gap at ~130 MHz, possibly also related to 
>>>>> flagging, and a discontinuity at ~180 MHz.  I was postulating the that 
>>>>> latter might be due to different 30.72 MHz chunks being observed on 
>>>>> different nights.  The explanation for the overall larger values at 
>>>>> higher frequencies seems totally plausible.
>>>>> 
>>>>> On Wed, Oct 7, 2015 at 7:50 AM, Eloy de Lera Acedo <[email protected] 
>>>>> <mailto:[email protected]>> wrote:
>>>>> Hi John, 
>>>>> 
>>>>> Do you mean the feature at 100 MHz? That one is not predicted by the 
>>>>> modeling and indeed it is very strange because it does not show up in the 
>>>>> other polarisation. We will check with our Australian colleagues, see 
>>>>> what they think. That should not be there.  
>>>>> 
>>>>> The reason why the A/T measurements give larger values than the model can 
>>>>> be related to this:
>>>>> "For Hydra A, our model predicts the flux density at 300 MHz to be ∼1/3 
>>>>> that at 74 MHz. However, detailed measurements specifically targeting 
>>>>> Hydra A at 74 and 330 MHz [35] show that the flux density of the compact 
>>>>> central region reduces less steeply with frequency, and would be only 
>>>>> ∼1/2 that at 74 MHz. Unlike the surrounding diffuse emission, the compact 
>>>>> central region is not spatially filtered at the higher frequencies, thus 
>>>>> our modelled "known" flux density causes an underestimate of the actual 
>>>>> flux density, resulting in a higher than actual A/T . In future 
>>>>> processing of measurement, we expect significant convergence as the 
>>>>> models of calibrator sources are improved for the MWA and other wide-band 
>>>>> low-frequency telescopes." 
>>>>> 
>>>>> Regarding A/T, in this paper (Fig. 15) you have our best estimates for 
>>>>> A/T calculation for SKA before re-baselining (before halving the number 
>>>>> of elements).
>>>>> 
>>>>> http://rd.springer.com/article/10.1007/s10686-015-9439-0 
>>>>> <http://rd.springer.com/article/10.1007/s10686-015-9439-0>
>>>>> 
>>>>> Cheers,
>>>>> Eloy.
>>>>> 
>>>>>> On 7 Oct 2015, at 15:31, Jonathan Pober <[email protected] 
>>>>>> <mailto:[email protected]>> wrote:
>>>>>> 
>>>>>> I'm not used to thinking about A/T this early in the morning, but are 
>>>>>> Figures 14 and 15 what we've always wanted to see from the SKA?  
>>>>>> Measured (and simulated) frequency responses?
>>>>>> 
>>>>>> http://arxiv.org/abs/1510.01515 <http://arxiv.org/abs/1510.01515>
>>>>>> 
>>>>>> (There seems to be a prevalance of 15's here, which numerology.com 
>>>>>> <http://numerology.com/> tells me corresponds to "Loving, forgiving, 
>>>>>> tolerant" -- for those of you perhaps looking for guidance in your 
>>>>>> relation to the SKA.)
>>>>>> 
>>>>>> In Figure 14 there are a few sharp features beyond what the model 
>>>>>> predicts.  I need to read the paper in more detail to figure out exactly 
>>>>>> what the measurement is.  Those could be due to sidelobes from other 
>>>>>> sources affecting their spatial filter, although they are using the full 
>>>>>> 128 tiles of the MWA for cross correlation, so the PSF should be good.  
>>>>>> They're also stitching together the band from several 30.72 MHz chunks, 
>>>>>> so those could be artifacts at band edges.  
>>>>>> 
>>>>>> Maybe Eloy and/or Nima have some thoughts on these results?  Very nice 
>>>>>> to see this work taking place and moving ahead.
>>>>>> 
>>>>>> Cheers,
>>>>>> 
>>>>>> Jonnie
>>>>> 
>>>>> -- 
>>>>> Dr. Eloy de Lera Acedo
>>>>> 
>>>>> Senior Research Associate
>>>>> Astrophysics Group
>>>>> Cavendish Laboratory
>>>>> University of Cambridge
>>>>> JJ Thomson Avenue
>>>>> Cambridge CB3 0HE
>>>>> 
>>>>> Teaching Associate and Bye-Fellow
>>>>> Downing College 
>>>>> Regent Street 
>>>>> Cambridge CB2 1DQ
>>>>> 
>>>>> Telephone: (+44) (0)1223 (3)37365
>>>>> Fax: (+44) (0)1223 337563 <tel:%28%2B44%29%20%280%291223%20337563>
>>>>> Email: [email protected] <mailto:[email protected]>
>>>>> Webpage: http://eloydeleraacedo.weebly.com 
>>>>> <http://eloydeleraacedo.weebly.com/>
>>>>> 
>>>> 
>>>> -- 
>>>> Dr. Eloy de Lera Acedo
>>>> 
>>>> Senior Research Associate
>>>> Astrophysics Group
>>>> Cavendish Laboratory
>>>> University of Cambridge
>>>> JJ Thomson Avenue
>>>> Cambridge CB3 0HE
>>>> 
>>>> Teaching Associate and Bye-Fellow
>>>> Downing College 
>>>> Regent Street 
>>>> Cambridge CB2 1DQ
>>>> 
>>>> Telephone: (+44) (0)1223 (3)37365
>>>> Fax: (+44) (0)1223 337563 <tel:%28%2B44%29%20%280%291223%20337563>
>>>> Email: [email protected] <mailto:[email protected]>
>>>> Webpage: http://eloydeleraacedo.weebly.com 
>>>> <http://eloydeleraacedo.weebly.com/>
>>> 
>> 
>> -- 
>> Dr. Eloy de Lera Acedo
>> 
>> Senior Research Associate
>> Astrophysics Group
>> Cavendish Laboratory
>> University of Cambridge
>> JJ Thomson Avenue
>> Cambridge CB3 0HE
>> 
>> Teaching Associate and Bye-Fellow
>> Downing College 
>> Regent Street 
>> Cambridge CB2 1DQ
>> 
>> Telephone: (+44) (0)1223 (3)37365
>> Fax: (+44) (0)1223 337563 <tel:%28%2B44%29%20%280%291223%20337563>
>> Email: [email protected] <mailto:[email protected]>
>> Webpage: http://eloydeleraacedo.weebly.com 
>> <http://eloydeleraacedo.weebly.com/>
>> 
>> 
>> 
>> -- 
>> 
>> National Science Foundation Fellow
>> Arizona State University
>> School of Earth and Space Exploration
>> Low Frequency Cosmology
>> Phone:           (505) 500 4521 <tel:%28505%29%20500%204521>
>> Homepage:     http://loco.lab.asu.edu/danny_jacobs/ 
>> <http://loco.lab.asu.edu/danny_jacobs/>
> -- 
> Dr. Eloy de Lera Acedo
> 
> Senior Research Associate
> Astrophysics Group
> Cavendish Laboratory
> University of Cambridge
> JJ Thomson Avenue
> Cambridge CB3 0HE
> 
> Teaching Associate and Bye-Fellow
> Downing College 
> Regent Street 
> Cambridge CB2 1DQ
> 
> Telephone: (+44) (0)1223 (3)37365
> Fax: (+44) (0)1223 337563 <tel:%28%2B44%29%20%280%291223%20337563>
> Email: [email protected] <mailto:[email protected]>
> Webpage: http://eloydeleraacedo.weebly.com 
> <http://eloydeleraacedo.weebly.com/>
> 
> 
> 
> -- 
> 
> National Science Foundation Fellow
> Arizona State University
> School of Earth and Space Exploration
> Low Frequency Cosmology
> Phone:           (505) 500 4521
> Homepage:     http://loco.lab.asu.edu/danny_jacobs/ 
> <http://loco.lab.asu.edu/danny_jacobs/>
-- 
Dr. Eloy de Lera Acedo

Senior Research Associate
Astrophysics Group
Cavendish Laboratory
University of Cambridge
JJ Thomson Avenue
Cambridge CB3 0HE

Teaching Associate and Bye-Fellow
Downing College 
Regent Street 
Cambridge CB2 1DQ

Telephone: (+44) (0)1223 (3)37365
Fax: (+44) (0)1223 337563
Email: [email protected]
Webpage: http://eloydeleraacedo.weebly.com <http://eloydeleraacedo.weebly.com/>

Reply via email to