Hi Zaki,
I just finished reading through your draft -- wow, this is an impressive
amount of work! There are a lot of stylistic/grammatical errors and a few
mis-citations that hopefully you'll be able to catch. This was a
relatively quick read-through so I can write Section 2 of my paper, so I've
tried to focus on major things. These comments are based on the draft with
git hash b7307b9bb12++, so apologies if some of them have been addressed.
I know there are a lot of them, but it's a long paper.
1. There needs to be a comma between David Moore and my name, otherwise
we'll get another D. C. Pober situation.
2. The sentence beginning "Under the assumption..." and the following
sentence should be removed from the abstract. The support for these
statements comes from my analysis, which is not in this paper. (Also, for
when you're discussing this result, the heating comes from galaxies, not
quasars.)
3. You use "Precision Array for Probing the Epoch of Reionization" and
"Precision Array to Probe the Epoch of Reionization" interchangeably. Best
to pick one. I prefer the former.
4. Figure 2 says yellow boxes are signal-loss free steps. Assuming
everything is shaded correctly, some of the lossy steps are very subtle
and/or perplexing. Delay transform causes signal loss? Isn't it just a
Fourier transform?
5. Last sentence of Figure 4 caption: "Residuals are consistent with
thermal noise." Is this calculated, or just an assertion? I would think
the fact that the mean chi^2 is 2 means the residuals are in general not
consistent with thermal noise.
6. Figure 6: why are the sidelobes of Fornax so different than the
sidelobes of Pictor?
7. I said I wasn't going to comment on typos, but the word
"polariztionats" just below equation 4 is too awesome to not point out.
8. The top labels on Figure 8 are not in the same order as the caption.
Also the caption says 4 orders of magnitude foreground suppression, the
text says 5.
9. 4th paragraph before section 3.4 starts: The argument that you can do
10 minute x-talk removal after the wideband delay clean is... unsettling.
Foregrounds/EoR signal may not dominate the average, but compared to all
the detailed signal loss calculations, this section just looks sloppy.
10. Second paragraph in 3.4: the rotation rate of the earth is constant;
I think you mean the rotation rate of the sky. In general, this section is
not very clear. Is Adrian going to have a fringe rate filtering paper out
next week to cite? If not, this needs work.
11. 2 full paragraph on page 8: "The action of the fringe rate filter is
get a nearly optimally weighted sample..." -- why nearly? Optimal has a
specific meaning, and you use it often elsewhere in the paper, so what is
"nearly" optimal here?
12. I'm not sure what "known issue" needs to be fixed in Figure 11, but
the units on the y axis can't be both K^2 and arbitrary units. Also, the
caption and the text disagree about the cause of the 2 day spike.
13. Just after equation 7: The fact that applying a median filter on the
lst bin variance requires a correction factor for estimating Tsys should be
in the text (right now its in a parenthetical side conversation). Also, do
you know what you'd get for Tsys without applying the median filter? If
that's much bigger than 550, it's an interesting comment on the necessity
of the median filter.
14. Paragraph after equation 10: how is the foreground bias accounted
for by cross correlating different days and baselines? If the days and
baselines are truly redundant, the foreground bias shouldn't average down.
Anything that does is some sort of systematic error.
15. In the discussion of M before equation 14, you say M = 1/F gives
large error bars and "hence the estimate is not very well localized."
That's not true, right? This *is* the best localized estimate, just with
crappy error bars. You then say that both M = F and M = D produce the
smallest possible error bars. This section needs work.
16. Just before equation 18, you say the empirical estimate of the
covariance matrix is due to "our limited universe." Are we really cosmic
variance limited in estimating C? Or could we get more samples with more
data/observations?
17. Are we going to explain why the wide band delay filter lowers the
eigenvalues of all eigenmodes and not just the foreground ones? Can you at
least explain it to me?
18. Just before equation 20, you mention that you're not using
baseline-to-baseline covariances. You should probably stress this more,
since its such a big change compared to P14.
19. 2 paragraphs before Table 1: "For each integration and frequency
channel, we draw a random sample from this distribution and assign it to
every baseline used, drawing a new sample for every integration only.
Therefore, every baseline agrees on the signal for each integration and
each frequency channel." Does this mean your EoR signal is completely
uncorrelated in time? Is this realistic? If you're optimally fringe rate
filtered, I suppose every sample is uncorrelated in time with respect to
sky signal... but I'm not 100% comfortable with that.
20. Last sentence before section 6. The mean and the median have to be
the same statistic for a Gaussian distributed signal. I thought what we
were discussing was that they were different for a Rayleigh distribution,
which is what a foreground biased power spectrum would be.
21. What changed in the noise curve relative to Monday? It looks a lot
closer to the data, but it's not so good to have most of your data points
below the noise, either.
22. The explanation of the k = 0.3 detection as polarization leakage is
just at the hairy edge of David's latest result. So I wouldn't just glibly
say it's "consistent."
23. I think you're selling me a little short in the paragraph before
6.2. My paper doesn't have a "detailed examination" of the results about
heating from the observed galaxies. That is the result of my paper. You
should absolutely discuss the results of my paper here, but it should be
clearer that they do not follow from the analysis prior. The language you
use in the conclusions is better.
24. In the first paragraph of 6.3, you should use the PAPER sensitivity
numbers that were updated in the Erratum:
http://adsabs.harvard.edu/abs/2014ApJ...788...96P. So 2.02 and 4.82,
rather than 1.65 and 8.86.
Happy long weekend -- hope all of it's not spent writing.
Jonnie
On Fri, Feb 13, 2015 at 8:47 AM, Gianni Bernardi <[email protected]
> wrote:
> Hi Zaki,
> thanks for sending this around! I attach my comments up to pag. 14 (as
> far as I've managed to reach). Hope they're helpful. If you have questions
> about them, feel free to ask. I'll send you the comments (if any) on the
> remaining part by Monday. It looks really good!
> Cheers
>
> Gianni
>
> On Mon, Feb 9, 2015 at 4:43 AM, Zaki Ali <[email protected]> wrote:
>
>> Hi All,
>>
>> Thought I would send around a draft before tomorrows meeting.
>> Still a little rough on the language, but everything should be there.
>> See you all tomorrow.
>>
>> You can either get it with (~10 GB)
>> git clone [email protected]:zakiali/psa64_pspec.git
>>
>> Or see the attached pdf.
>>
>> Zaki
>>
>>
>>
>>
>>
>>
>
