-
Notifications
You must be signed in to change notification settings - Fork 0
/
descriptions
79 lines (63 loc) · 7.17 KB
/
descriptions
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
-------------------------------------------------------------------------------------
Hi folks, here the pointed observations. I haven't married them with the simultaneous ASM data yet, but I am getting there. As always, a lot of little unanticipated problems on the way. Anyway, fits file here
http://www.sternwarte.uni-erlangen.de/~grinberg/all_data_for_ml.fits
with the following entries:
-block = identified
-tstart = start time in MJD [days]
-tstop = end time in MJD [days]
-orbitalphase = orbital phase of the system (I used the midpoint of the observation; observations are usually 30 min long, the orbital period is 5.6 days, so this is an OK approximation)
-smoothorbitalphase = a continuous but *not* smooth parametrization of the orbital phase
-gamma = spectral slope / photon index from spectral fits; equivalent to: state label: Gamma_1 <= 2.0 hard state, 2.0 < Gamma_1 <= 2.5 intermediate state, 2.5 <= soft state
-nh = absorption / equivalent hydrogen density [cm^-2]
-lag14 = average timelag in 3.2-10 Hz range (9.4-15 keV vs. 2.1-4.5 keV)
-cof14 = average coherence in 3.2-10 Hz range (9.4-15 keV vs. 2.1-4.5 keV)
-rms1 = rms in 0.125–256 Hz range in the 2.1-4.5 keV energy band
-rms2 = rms in 0.125–256 Hz range in the 4.5-5.7 keV energy band
-rms3=rms in 0.125–256 Hz range in the 5.7-9.4 keV energy band
-rms4=rms in 0.125–256 Hz range in the 9.4-15 keV energy band
-avg1=average countrate in 2.1-4.5 keV energy band
-avg2=average countrate in 4.5-5.7 keV energy band
-avg3=average countrate in 5.7-9.4 keV energy band
-avg4=average countrate in 9.4-15 keV energy band
-en_lo=each entry is an array of length 78 containing the lower bounds of spectral energy bins
-en_hi=each entry is an array of length 78 containing the upper bounds of spectral energy bins
-flux= each entry is an array of length 78 containing the flux values of the spectrum in the corresponding spectral energy bin [nuFnu]
-flux_err=each entry is an array of length 78 containing the error of flux value of the spectrum in the corresponding spectral energy bin [nuFnu]
Not all blocks have corresponding variability characteristics (lag, coherence, rms, average count rate) - in such cases, the values are set to NAN. If it is not easy for you to filter for that, I can easily re-do an extra table, only with data that has all the info. We will need all the rms or average values to characterize the data - I do not know which combination is the best however, that's something that I hope the data will be able to tell us. I can also give more parameters (say lags and coherence function values between different energy bands), but I feel this would be even more redundant. Daniela, you did look into which of the parameters looked to be most useful, didn't you? I think it would be easiest if you ask question about the data than if I just smother you with information :)
Cheers, Victoria
------------------------------------------------------------------------------------------
Hi folks, here the ridiculously large ascii file with all the all sky monitor data. Some remarks on this - partly for me, to have a second place where the notes are made, next to my notebook.
The columns are named in the files, just to be on the safe side:
1. time in days expressed in MJD (modified Julian date)
2. orbital phase calculated with the ephemeris of Gies 2008; this may become important as there is a known orbital variability (orbital period of 5.6 days)
3. total counts in the 1.5-12 keV band
4. error in the 1.5-12 keV band
5. counts in the 1.5-3 keV band (band A)
6. error in the 1.5-3 keV band (band A)
7. counts in the 3-5 keV band (band B)
8. error in the 3-5 keV band (band B)
9. counts in the 5-12 keV band (band C)
10. error in the 5-12 keV band (band C)
Note that that the sum of columns 5,7 and 9 is usually not the same as the value in the column 3. (I can explain why but I am not sure you find this interesting?)
I applied two filters to the original data:
1. measurements where counts in any of the three bands (A,B, or C) were smaller or equal zero were taken out - these are clear background oversubstractions. (You make a measurement, you estimate the background and substract one from the other - if the results is below zero, you clearly oversubstracted. This happens always because of the more or less Poission distribution if the signal, we are counting individual photons here.)
2. I only use data before MJD 55200 as the ASM (all sky monitor) instrument started deteriorating afterwards. So we effectively have 14 and not full 16 (lifetime of this the RXTE satellite) years of data. Cheers, Victoria
----------------------------------------------------------------------------------------------
>> This data has no labels, right?
Yup!
>> I remember you showed some fraction of these measurements in a plot. Which ones were those? B / total count?
C/A vs. total. It has to be the fraction as what we are looking after is, in terms of physics of the systems, how the shape of the spectrum behaves vs. how bright the source is.
>> It looks like the data made up of the three bands is pretty much two-dimensional, with B = 0.3567 * A + 0.4270 * C + 0.302 (R2 = 0.938)
>> Is that expected / plausible?
Yes. I am not sure about the values but what I think that what you are seeing here (this is me interpreting your results, I may be wrong) is the fact that the main spectral change is the spectrum getting more or less steep. (See attachment for my attempt to visualize what I mean.)
>> I did a 3d scatterplot and there is only very little variance in the other direction.
Can you do one for total vs. C/A vs. C/B? This may make things more clear.
>> I attached some heatmaps. It looks somewhat like two clusters, but one is much much more dense (i.e. many more points) than the other. You need to use log-density for the second cluster to even be visible.
I do think you are catching the soft state here - I am not 100% sure, I am more used to the C/A vs. total way to see it.
>> I also checked how it looks on log rate A vs log rate B which also looks kind of interesting.
I think it's the same effect you are catching in A vs. B as A vs. C.
>> Is it expected that one of the clusters has so many more points? That might make it hard to fit a model, but it's probably still possible. Do you have any estimate on how much larger the one cluster is than the other? I'll try to see what I'd estimate.
Yeah, that's expected. We will generally have the problem that Cyg X-1, during the lifetime of RXTE, spent most of its time in the soft state. I can do a short estimation tomorrow, when it's less late. Very roughly, without actually doing the numbers -- 1:9 for how many points are in one cluster vs. the other, rather leaning towards even more imbalance. Which also reminds me that I should perhaps also send the paper this whole idea is based on around. No necessity to read it, but I may refer to plots there. (Also: they screwed up my Figure 5 - the y-label should be "total counts"; it was there and they cut it off and I was a naive first-time-author who did not check whether they broke my figures when I got the copyedited version.)
--------------------------------------------- repos ---------------------
https://github.com/amueller/cyg-x-1
https://github.com/dhuppenkothen/cygx1class