Nova Del 2013 Photometry

Baader filter has a different response than Astrodon filter.

Carlos Tapia, an optical specialist at the Universidad Complutense in Madrid, Spain, who works in the LICA optical components testing laboratory, has tested a wide sample of filters, some of them photometric. Test curves can be reached at

http://carlostapia.es/curvas_filtros/revisiones.html

The response curve for Baader V filter can be seen here:

http://carlostapia.es/curvas_filtros/Baader_V_Johnson.html

And Astrodon V filter (two identical filters were tested, one of them is mine):

http://carlostapia.es/curvas_filtros/Astrodon_V_Johnson.html

It can be clearly seen that Baader has a "tail" in the red part of its response. Optec V Bessell seems to have a similar curve.

Miguel and Arne, thanks for the useful information. This makes a lot clear. In the future I want to do more of this kind of work so I will consider getting an astrodon B and V filter sometime. I just became an AAVSO member and love the vphot program btw. I just wondered if there is any idea about making a stand alone program for this? My files are 32mb and uploading dozens of them doesn't make me happy :) But maybe that's more for another topic. 

Arne: I'm curious if my data for the rest is ok. Thanks again for your effort starting this topic and helping.

Just added another time series starting around 2456544.5 .

Here we get a nice overlap for three observers: MZK, LDJ and AHM, see LCG graph:

http://goo.gl/V2U3zt

I highlighted MZK since his data points are lying "in the middle". I am not sure what happened to WGR, his data (top series) are untypically higher and show some "outbursts"... Did he, Gary, try a different filter?

My data are the large cloud underneath WGR's series. I created some (ugly) Excel graphs, that are kind of interesting (at least for me), see attachments:

NovaDel2013_JD2456544.6_Vmag+Cmag+Kmag_20130909.png is an overview of the whole run, displaying V, Comp and Check star. In the middle of the run some faint clouds came through (air plane trails?), but they did not affect the Vmag (as they shouldn't). 

NovaDel2013_JD2456544.6_Vmag&Kmag-Cmag_20130909.png shows zoomed in Vmag and the difference Kmag-Cmag. 

This was a mammoth run for me: It was the first time that I could monitor the data acquisition from inside my home. I was targeting 6 hours, but after 5 hrs. I admit I fell asleep in front of the monitor... How do you guys/gals stay up all night?

When I woke up (I had dozed off for only a few minutes) and checked the monitor I noticed immediately that something was wrong - the star formation had changed! Something seemed to eat our nice nova... Time warp, neutron star, degenerate matter...???  Lateron I followed up and checked the saved single images (more than 600, crazy!).

Call it "luck of the newcomer", what I found you can see in the last attachment...

Is there a way to report the finding through WebObs?

Cheers,

Helmar

So basically, please correct me if I'm wrong, the difference is caused by the influence of the H-a and H-b lines. Using the Baader filter I catch part of these lines and get another magnitude.

So that's a handicap I have to deal with. I was wondering if there will be any way to correct for this somehow. On the other hand I was thinking if this will not have an extra scientific value. My time series will be different of the time-series by the other filters because of the H-a influence. So basically the difference is a measure for the strength of the H-a signal? Combined with spectra I can imagine this adds some info, or am I thinking the wrong way? Probably the spread of the measurements is too high to measure this, but still, the principle is interesting I think...

André

There are several problems here that could be solved.

1/ We should use filters from only ONE  manufacturer Certified by AAVSO.

2/ We should use the SAME Comp/Check stars.(Rekommendation stars)

3/ Automation could do ALOT like looking inte FITS header and Calculate saturation levels from
CCD Camera spec vs time and just kill images with saturated stars.
Or not allowing them to be measured if star is saturated.

3/ AAVSO sequences should NOT contain doubious stars like star 80( 000-BLC-955 ) wich has a lot of 
 fainter stars in near aperture size ring.
It shouldnt even be there and have I never used it as comp or Check star,

I also raise doubt of using BIG aperture photometry on bright objects like Nova Del.
I can understand the temtation but how abt the quality of sub second photometry??

For the guys reporting multiple filter data (transformated) i dont have much to say except that
we should use the same filter ( manufacturer and AAVSO Certified)

I suppose that its hard to Certify filters bec the sellers and manafacturers would rise hell but 
if not ,,how can we ever have "perfect" data.

Else there could be ex a V(filter X) option instead of just plain V for the reports.

/Pierre

Roger is correct, in that the Rc or DSLR-R channels are heavily influenced by Halpha, and so give a clue as to how to correct the V-band measures to account for any red-tail-Halpha contribution.  Here are a couple of papers to read that discuss red leaks, filter responses, etc.:

Munari and Moretti, 2012BaltA..21...22M

Munari et al.,  2013NewA...20...30M

Basically, if stars behave reasonably smoothly through a filter bandpass, then it is straightforward to transform the data onto a standard system.  If there are sharp features, such as emission lines, it gets far more difficult.

One method of correcting for the red wing of some filters, like the Baader V, is to observe at V, Rc, Ic.  Rc is affected by Halpha; Ic is not.  If you convert the magnitudes into fluxes, you can see the increase in brightness at Rc, compared to a straight line between V and Ic (sorta forming the continuum).  That tells you the amount of the increase at that date, but does not tell you how to correct the result.  You would have to do this for several objects with differing Halpha contribution to create the data set for, say, a linear fit.  I think it could be done, but I bet it is not simple.  My expectation is that we can adjust the zero point between observers as long as there is sufficient overlap.

As Pierre suggests, we could "standardize" on filters from one vendor.  However, for now, if your data appears to be brighter than the typical observer, then perhaps include a note indicating what filter vendor you are using or any other equipment details that might be useful to know.

Arne

I've uploaded ~600 BVRI datasets from BSM_Berry (on the roof of HQ) for the UT nights of 130906, 130907 and 130909.  These are roughly transformed; I don't have good coefficients yet.  For the target star itself, transformation makes no difference, but it does for the comparison stars.  The BSM data overlay some observers but not others; they indicate the fainter observations are the closest to the standard system.  I will redo these points once I get good coefficients, and also bin them to improve signal/noise and reduce the number of submitted observations.  There are also an hour's worth of BSM_Hamren data (fully transformed) for 130908.  All of these points are under the HQA obscode.

Gary's observations from 130909 are the recent outliers.  Not only are they 0.2mag too bright, but they exhibit flaring activity which is not seen in other observers at the same time.  I think those images should be looked at carefully.  I see similar "flares" on his 130825 data too.

That said, I'm seeing nightly variations.  It is a little suspicious that they peak around transit, but they don't peak exactly at transit and have the same shape and amplitude in all filters.  Since other observers are seeing nightly variations (see for example 130906, where 5 separate observers show a fading trend), I'm inclined to believe that these variations exist - but when you see a new feature, it is always suspect and needs to be studied carefully.

Some other observations:

PXR 2456532 has a dozen points from 0.35632 to 0.35688, an interval of about a minute.  These are given with an uncertainty of 0.02mag, and with a peak-peak scatter of 6.488 - 6.415 = 0.073mag.  It would be far better to give a single datapoint under these circumstances.

There are several observers using unfiltered cameras.  I really recommend that you observe some other target.  Novae have all kinds of structure, like emission lines, that really mess up any kind of magnitude estimation with an unfiltered system.  Your magnitudes are usually too bright, and can't be correlated with anyone else's.  There are plenty of stars where unfiltered photometry is useful; just not bright novae.

The light curve is trending downward at about 0.05mag per day, after its 4-day pause.  We are 25 days after the peak, and the nova is still bright!

Arne

Hi all :)

I'm Carlos Tapia, the builder of the filter database. I'm looking to open the transmission values of some filters that are important for investigation (photometric only), like this work with the nova.

One comment to de Baader V filter, is physicaly composed by a stack of two filters, like the original Johnson photometric filters, use the same glasses. The new filters like Astrodon are interference filters.

I will be delighted to help with whatever AAVSO users need with filters.

Kind regards

I'm slowly working through the submitted images; please excuse my speed!  Miguel contributed a set of 40 images that he took on August 25, 2013UT.  There were 10 images in each of BVRI, taken with a 200m f/8 R/C telescope and a QHY9 camera (KAF8300).  He purposely defocused the stars, which is fine - the peak pixel value in the images is under the saturation limit for the 8300.

In fact, Miguel used a 3x3 binning mode for this camera.  I hesitate using such a mode with amateur cameras, as the readout node can really only handle about 2x full well electrons.  Since the KAF8300 has a full well depth of 25K electrons, this means that you can only expose to about 50K electrons.  Since 3x3 means 9 native pixels contribute to the signal, each one can only be filled to about 5K electrons before the readout amplifier gets saturated.  You will get more dynamic range by either not binning, or at most binning 2x2.

That said, the images look fine.  The stars occupy about 25 binned pixels with their donuts, so the measurement aperture has to be at least that large in diameter.  Miguel used the 80 comparison, and that is a little troublesome, because it has two fainter companions to the E and NE that blend with the comp star when defocused.  Therefore, the comp star will appear a little too bright, and the variable will appear a little too dim.  However, I don't see any real problems with his processing.

Arne

Andre' contributed three images taken with his system, each one being a stack of five 4second exposures.  The telescope is a TEC140, using a QSI-583 camera (KAF8300; they are popular sensors!) and a Baader V filter.

The 28th and 30th had reasonable seeing, with fwhm around 3-4 pixels.  The 29th had poor seeing, with fwhm more like 11 pixels.  I hope you changed the aperture size for the poor seeing nights!  The peak ADU is fine; these chips are usually set up with a gain that digitizes the linear portion of the response curve.  The maximum ADU was about 36K on the 28th, when the variable was about V=6.7.

The main thing I noticed with these images is that V339 Del is very much off-center.  Was there a reason for this?  I don't have the master flat that was used, but I notice that the noise characteristics in the sky background are worse on the corners, so there is some vignetting in the system.  The Taks are usually pretty well baffled, so I don't expect scattered light, but I'd feel a lot better about this setup if the target were closer to the center of the field.

The other concern is that I measure V=6.71 from the 8/28 image, yet you report V=6.5 or so if I read the LCG properly.  I was measuring with respect to the 80 comp star, and from your stack.  Your reported photometry may be the individual 5 images, but even so, this seems like a big difference twixt your measure and my quick-and-dirty iraf measure.  On the 29th and 30th, we agree pretty well.

We've already discussed the possible offset due to the red wing of the Baader filter, and that is probably the major contributer to Andre's offset from other observers.

Arne

Zap is a wonderful tool; thanks, Sara!  Zapper is very similar for the non-staff to use; you should try it out.  In particular, today we will talk mostly about 2456521 and 2456522, around the peak of the outburst.

If you want to see what you should be getting with CCD systems, look at the data from PGD on 2456521.  He was using an Optec SSP-3 single-channel photometer, but we have a million channels - why can't we do this good?  The data are precise, to within a millimag or so; there is no question of a fade during the time series.  The only thing missing that I see in the observations is the uncertainty measure. In contrast, look at the slightly earlier data on that night, and on the following night, from DTTA.  This observer is doing the same filters (BVR), but has scatter 10x what the PEP observer was doing.  This is what I'm trying to achieve: if all CCD observations looked like the PEP observations, researchers would be ecstatic and would clamour for AAVSO data.

Back to DTTA's data:  the observer quotes an uncertainty from 0.005mag to 0.05mag per point.  Yet, the observations have peak-peak amplitude of 5.12-4.64 = 0.48magnitudes.  This looks like saturation to me.  The observations are also untransformed; BV basically follow PGD pretty well, but R is too bright (probably even worse saturation). Another clue is that DTTA also observed at U-band (very good!). The U-band observations have far smaller scatter, as U requires much longer exposures than the other bands and therefore is less likely to saturate. So several comments: (1) the error analysis is wrong; (2) the images are likely saturated; (3) with this poor of a dataset, there is no reason to be reporting with this cadence; (4) such datasets obscure the really good ones, like from PGD, when using the Light Curve Generator.  If DTTA is reading, I'd like to see a handful of the images from that night.

We've also marked at set of V-band measures on 2456523 from MFB as discrepant (they were 0.6mag too faint), which won't show up on LCG.  SAH has a BVRI set from 2456520 that look about 0.2mag faint, but only one set; the same for 2456521.  Later nights, SAH's measures agree pretty well with others.  I bet on saturation, which was corrected on later nights, but I'd like to see the images. HPIA and LMZ also have single V-band estimates on that date that look too faint. NRNA has a small group of BVR measurements that should have been averaged and a single point submitted.  NRNA's error estimate is also far too small; the I-band measures, for example, span 5.00-5.12, yet are reported with 0.0085 error.

Now, the question is what to do with data like the BVR measures from DTTA.  We usually go back to the observer and ask them to inspect their images and see if there is a problem.  Perhaps the observer was imaging through clouds, and so a larger uncertainty should be applied.  Perhaps the images are saturated, in which case the observer should remove them from the database (you can't recover good data from saturated images, especially if the sensor is ABG).  It is hard for AAVSO staff to know whether a dataset is good or bad, so we have to rely on the observers to carefully examine their own data before submission (or fix it after submission).  That "fix it after submission" is important - you can go back and correct your data, and that is far better for the researcher (and less embarrassing for you) than leaving discrepant data in the database. The more our data looks like PGD, the happier I will be!

Arne

Neil sent me a set of images from 2013-08-30.  He has a unique system, using a 50mm camera lens along with a filter wheel and an ST-7 ABG camera.  He was running the camera at -16C, with 60second exposures at V.  BRI are mostly in focus; V is out of focus due to the optical configuration.  Neil, what focal ratio do you run this system at, and whose filters?  Most 50mm lenses are pretty fast, and most interference filters don't like fast beams.

His data pretty much fall on the generic light curve for the nova.  However, I have a few comments about the images.  He takes multiple images and stacks them before submission - excellent.  However, the variable is near one edge rather than being centered.  He has enough field of view that there are plenty of comp stars with the target centered, so centering is a far better situation.  The system also has vignetting, so centering reduces any residual effect.  The out-of-focus V images are interesting; they have a reasonably sharp core along with a wide "base".  At least, these will be easier to centroid with most software packages.

Besides centering, the other thing that bothers me are the flats.  While Neil keeps the peak ADU in each science image low enough that the ABG effect does not impact the linearity of the photometry, the flats are exposed to a much higher level.  I calculate the peak ADU in the V flat to be about 51K, for example.  I would be very careful and determine that this is still within the linearity range of the camera before exposing to this level for your flats.  It is usually better to take more flats at about 30K ADU to reach the same signal/noise, than to take fewer high-level flats.

I think this is all of the images that have been sent to me.  I've requested others, but those observers have not responded.  If I've missed any, or anyone wants a critique, let me know.

Arne

To date, I've looked at the light curve daily and checked for obvious discrepancies.  The issues that I've seen with the datasets:

- saturation.  This is becoming less of an issue now that the variable has faded to V=7.5, but makes some of the early observations questionable.

- too many submitted datapoints.  If you take a quick grouping of images on one night, you should be reporting the average and standard deviation, not every single image.  I know it is extra work, but the researchers will thank you.

- incorrect settings.  A couple of observers used the wrong filter; others placed the target away from the center of the field.  Some used too small an aperture; others had problems with one comp star but not another.

- not looking at your data.  Surely you can spend a few minutes to see how your measures compare with others after submission.  That can catch some obvious errors very quickly.  Those who are observing defocused will have more difficulty in looking at the images, because donuts are not nice clean stellar profiles.  They are great ways to see how poor your collimation is, though!

- submitting with improper uncertainties, or times with millisecond precision.  Observing unfiltered.  Learn the limits of your equipment.

- no transformation.  While Halpha does affect the later photometry as the continuum fades, you can get much closer to the standard system if you transform.  The nova is basically colorless now (but not earlier), but the comp stars are much redder, and applying transformation corrects for their color.

- differences in filters.  As shown in this thread, Halpha is the culprit with most novae.  It is bright and wide, and contributes quite a bit of flux.  Filters from different vendors have different passband shapes, with some having more throughput at Halpha than others.  This will cause a basic offset between observers, even after transformation.

I would also like to suggest that observers consider doing an all-night time series, as this makes it MUCH easier to see how your data overlays with other observers.  Finally, I notice that the number of observations is decreasing.  People are getting bored, or there is more bad weather this week!  The nova is still BRIGHT; even at 7th magnitude, it is the brightest nova we've had in years.  Get out, keep monitoring, see how long you can follow it.  I bet everyone can still get good signal/noise 6 months from now - wouldn't it be wonderful to have a high-accuracy light curve of a nova that is months long?

Arne

First, I want to echo what Doug said: we're working together to improve the photometry submitted to the AAVSO.  Many of you do just fine; others just needed slight modifications to their techniques, or needed to devote more time to examining their results.  These are not huge changes, but reap huge rewards.  I appreciate your effort!

As for scattered light: the problem here is in the flats, not in the science frames.  If the moon is nearby a science field, you will get scattered light into your image, often in the form of a gradient or different illumination of dust.  This is ok; it is an additive feature, as suggested by Roger, and normal aperture photometry gets rid of it.  However, when scattered light is in the flat, then it modifies the assumption of uniform illumation at the front of the telescope.  This scattered light contribution then gets divided into every science frame, and modifies the photometry.  A large part of my CCD School is devoted to these kinds of systematic errors and how to get rid of them.

Likewise, optical distortion modifies the basic assumption that all pixels see the same amount of sky.  This also changes the photometry.  If you have coma, or your chip is larger than your corrected field of view, or if your corrector doesn't do a perfect job, photometry is compromised.  That is one of the reasons that I highly recommend that you put your target and comps either in the center of the field of view, or at least at the same radial distance from the center.

For the most part, the effects mentioned above are small issues, made even smaller if you are working with the center of a field of view.  The bigger effects are the ones I've mentioned before: saturation, wrong filters, scintillation, poor signal/noise, etc.  However, as you correct the easy ones, you progress to the harder problems - it is the nature of being a good photometrist.

Keep up the good work!  Lets continue to work on these stars and others to get the ultimate precision out of your imaging.

Arne

Next time when it is clear, I try NovaDel light curve. My scope is 30cmF4 and it collect too much light.

So I reduce my scope 30cmF4 to 11cmF11, with cardboard mask.

Drawback is still 1200mm focal length and it give me only two refstars 80 and 98, and my V-filter is Baader V-filter :(

See attachments.

That kind of mask should work fine.  A couple of suggestions:

- make sure that the aperture mask hole avoids the secondary mirror supports (spider).

- make sure that you take flats with the aperture mask on the telescope, as your flats will be very different with the mask than without the mask.

There are a couple of fainter stars between the 80 star and the nova that can be used as comparisons, or as checks if you want to use the 80/98 as a small ensemble.  These are about 10th magnitude and will become important as the nova fades.

Arne

A very different method for make precise photometric consist to take spectra scaled in physical unit (in erg/cm2/s/A, for example). Here the nova spectrum evolution from August 29 to Sept. 15 taken with a low cost spectrograph (Alpy 600) on a C11 telescope in suburban conditions:  

The same plot in log scale (note temporal evolution of features):

From these data it is possible to extract the magnitude in a photometric system (Bessel BVR):

The filters are of numerical nature (!). Details of the method (standard in many professionnal observatory) is given here:

http://www.astrosurf.com/buil/calibration2/absolute_calibration_en.htm

Contrary to appearances, the technique is not much more complex than the traditional photometry. And the content is very rich: the rich spectrum and accurate photometry, even when the spectrum is complex.

This is a new approach to photometry, very powerful!

Many amateurs produce spectra today...

http://www.astrosurf.com/aras/Aras_DataBase/Novae/Nova-Del-2013.htm

See also http://www.astrosurf.com/aras/novae/Nova2013Del.html

Why not you?

Christian  Buil

Christian has shown some good examples of converting calibrated spectra into equivalent broad-band photometric values.  This works remarkably well, as long as you pay attention to the details; using a known flux standard, having a photometric sky, widening the slit to ensure that seeing, etc. are not a limit.  However, those restrictions are far more specific than differential, broad-band photometry, where you gain a multiplexing advantage (measuring many stars across the field of view) while losing spectral resolution.  It is similar to the difference between PEP and CCD observing.

More people doing spectroscopy should try flux-calibrating their spectra.  It can be fun!

Arne

On 5/6 October (JD 2456571), I took some V band photoelectric measurements of the nova, using an SSP-3 photometer on a 24 inch telescope.  Observers SRIC and EEY overlapped with me, and I wanted to post this comparison of our results.  I appear to get significantly brighter values.

The four PEP observations were centered on JD 2456571.68229, .70694, .73333, and .75590, using 000-BCL-955 ("80") as the comparison star.  The PEPObs data reducer does not support the nova, so I crunched the numbers on my own (PEPObs reductions for other stars match those from my home cruncher).  The data were transformed using a B-V of 0.017 for the nova, yielding a delta B-V of -0.299.  Below, I have linearly interpolated the two SRIC magnitudes (also transformed) bracketing each of my measurements.

     JD fraction    interp. SRIC     CTOA      CTOA error    SRIC-CTOA

A:     .6823           9.770         9.698       0.001         0.072   

B:     .7069           9.812         9.721       0.005         0.091 

C:     .7333           9.840         9.752       0.006         0.088 

D:     .7559           9.884         9.796       0.003         0.088 

The B-V value for the nova was calculated by looking at the "best" B and V values SRIC took during this time.  I took pairs of successive B and V values, for which both measurements had errors less than 0.030 (arbitrary).  The individual B and V values were averaged separately, then subtracted.  Working backwards in time, there were 8 pairs, with B-V of 0.017.  Working forwards in time, there were 7 pairs and B-V of 0.015.  The difference of 0.002 of the two methods had no effect on my transformation results (I suppose I should have used 0.016, to be fussy). 

For those not familiar with PEP: a single observation takes place over about 15 minutes, and individual deflection times within the observation are only recorded to a one minute precision.  The following JD 2456571 pairs of SRIC data were used in the bracketing:

          JDs                       mags                errors

A:  .68012, .68771    9.767, 9.776     0.050, 0.053

B:  .70573, .71192    9.810, 9.819     0.133, 0.047

C:  .73200, .73453    9.846, 9.834     0.018, 0.030

D:  .75489, .75743    9.887, 9.880     0.026, 0.019

Because my clock had one minute precision, and it was set from a clock with one minute precision, there is room for slop in my times, but not nearly enough to explain the differences between SRIC and CTOA (the maximum rate of change between SRIC bracket pairs was 0.0033 magnitudes/min).

My check star was 000-BLC-945 ("98"), and its computed magnitudes (A..D) were 9.862, 9.872, 9.871, and 9.844.   The PEP algorithm does not transform the check star magnitude or correct for airmass difference relative to the comparison star.

Observer EEY was also active during this time, though his data are not transformed.  Given his rapid sampling (cadence under 1 minute) and my time uncertainties, I did not try interpolation of his magnitudes.  I selected and averaged three EEY values: the one closest in time to my own, and the ones immediately before and after.

      EEY center JD     EEY mean    CTOA      EEY-CTOA

A:       .68218         9.736       9.698      0.038

B:       .70690         9.756       9.721      0.035

C:       .73315         9.784       9.752      0.032

D:       .75583         9.834       9.796      0.038

EEY's errors were all 0.005.  If I turn off transformations in my reduction software, my own magnitudes brighten by about 0.014.

In short, I saw NV Del an average of 0.085 brighter than SRIC and 0.036 brighter than EEY.  I realize I have presented all these data rather scattershot; I would be happy to provide them in another format if that would help.

Tom Calderwood

Hello

Got a real nice run last night (27 images in each color, 60-120 seconds) in B, V, and Ic.  V and Ic turned out very nice.  The B values were 0.15 mag brighter than V.  I did not post to database as this looks weird.  Images look fine.  Taken in turn with V and Ic.  Nothing weird in images.  Its a new Astrodon filter with the good red blocking, supposedly.

Did anyone else get funny B results last night?

Gary

WGR