Egress Pause !

Interesting. How did you compute the period of the high frequency oscillation? It would be very interesting to have a high res (in time) lightcurve of EPSfrom several observers I guess. Physically speaking, what would we expect in terms of variance ? Between 2nd and 3rd contact, any variations caused by the disc (not the F star pulsation) would be caused (among other things) by some maybe "jagged" part of the disc getting in front of the F star disc while some other, also "jagged" part leaves the eclipse FOV at the otehr side. So you get some "noise" from both ends. After 3rd contact, you get noise only from one end of the F star disc, so to speak. Right? Maybe applying wavelet analysis to the light curve with a "bathtub"-lightcurve subtracted to remove the overall trend of the eclipse would be a good idea? *scratch head* CS Heinz

Dear all... prompted by Roger's posting, I'll be trying for some time series on eps Aur here. I approach this very cautiously though, as earlier Ihad a "false alarm", or a "false signal", on eps Aur. This seems like a good time to share the story. I had been doing time series on eps Aur with an old ST-7E (not microlensed) camera, and a 50 mm FLCanon SLR lens. I was using an equatorial mount, unguided. Polar alignment was good enough to keep the field on the chip, even for runs of many hours, but the stars wandered across the pixels. Then my mount died. As temporary measure, I moved my camera and lens and put them piggyback on a big scope, and I ran that mount autoguided. I started seeing beautiful signal! It had an amplitude of about 0.04 mag, and period about 0.04 days, similar to what Roger reports -- but that match is the purest of coincidences. My joy was shortlived (a second or two) because I saw the comp stars had similar signals. The universe was going crazy! Stellar pulsation everywhere! What it really meant was Ihad a nasty problem I had to solve. What it finally turned out to be, was sub-pixel-scale sensitivity variations on the CCD chip. On the Kodak E chips, one half of each pixel has a different spectral sensitivity than the other half. (Note that microlensing addresses this by funneling most of the light through one half of the pixel.) Even with defocusing (another story there, see below), I was pretty undersampled. So long as the stars drifted quickly across the pixels, the effect of the sensitivity variation was averaged out, or at least randomized so it didn't look like an astrophysical signal. But once I started autoguiding (now shooting 50 mm FL, on a big A-P mount, with guiding, so tracking was smooooooth), the PSFs moved across the pixels verrrrry slowly, even with my poor polar alignment. Attached is a graph showing the story. The bottom two panels show the fractional portion of the pixel position of one of the stars (eta), as reported by the photometry program. The fractional Y position within the pixel correlates with the "signal". Note there is no correction for airmass on that graph -- that's why all the light curves drop off with time So, what's my point? Not that Roger is seeing the same thing, because Ihaven't seen his data and don't know his setup. But just that we're pushing our equipment pretty hard, so subtle things can bite. What I've described is one that isn't usually on the list, but I think you'll agree that this put a beautiful, but completely spurious, "signal" in my data! I'm not a DSLR observer, I haven't been following the details of DSLR photometry, but I would think that because of the colour-pixel-masking used in those cameras, the "sub pixel" sensitivity variation could be even greater. What about defocusing? That is supposed to solve the problem, after all. Well, after this I looked closer at my PSFs and found that while they were getting fat, they were still rather pointy. I.e. the brightest pixel in a PSFstill got a lot more photons than the others. I'm still on the trail of fixing that -- the ideal defocused PSF would be quite flattish so equal amounts of light are hitting a bunch of neighbouring pixels. Check yours! Kind regards, Gary Billings, Rockyford, Alberta

Hi Garry, Thanks for your comment, your graph is very interesting. As I said my first approach during the past weeks during the egresshas beenit was a question of sky condition but I changed my judgement seeing repeatedly the other stars of the survey were not affected at all by that variance (my fully automated software records the data of 50 stars of each image of each session) This question of pixelisation is well known from DSLR photometrist, this is the reason why we use a large defocusing. The defocus used is such the PSF looks a flat disk, even sometime a donut (!)depending the lens aberrations. In fact I also use some star trail in addition (no equatorial motion) My software includes an adaptive reading apertureand for such evaluates the surface of the sensor that gets a detectable signal from the star, the result is usually something like 200 pixels. The star lightprofil is also recorded for further check. Our CMOS DSLRhave many small pixels (microlensed) compared to most CCD astro cameras and even if the Bayer structure is an issue with such many pixels it's not difficult to defocus on many of them. I should also mention DSLR have a low-pass (resolution)filter in front of the sensor that somewhat spread the lightto eliminate the Moiré effect due to the Bayer structure. As we would say in French this is a little bit "ceinture et bretelles" !! (belt and braces, you know, just tobe sure... ) Anyhow this interpretationfrom the results of a couple of days is an hypothesis ( I used the conditional) But the signal being well in phase for 7 days it seems to me the probability it's an artefact is not thathigh. I keep observing to confirm or not and results from others are surely welcome, please let me know yours. Yours truly, Roger

I happened to be shooting a time series on eps Aur last night, April 24/25. Not the greatest conditions, and airmass started at > 1.5 Equipment was a 135 mm lens at f/2.8, with an ND8 filter (reduces light by 8x, allowing longer exposure for reduced scintillation), and a V filter. ST-8XME, 25 sec exposures. Dark subtracted, flat fielded. Unguided. I've applied a correction for airmass (as you can see the night was not photometric as there are parallel wobbles in the light curves), but have *not* formed a differential, or transformation. I continued shooting out to airmass about 7, but the data is dodgy even before 3. The data is only tolerable until about JD 676.71, but that gives 0.71 - 0.643 = 0.067 days. Mag axis ticks are 0.05. By eyeball, I'd say if there is any hour-scale oscillation in this data, over that time interval, it is less than about 0.02 peak-peak. Does anyone else have a time series over the same interval? Skies looking worse tonight, unfortunately, Gary Billings

Hello Heinz, I did it very simple, as I have about a period for eachof three days I knew it was around one hour, next I tested the possible multiples around that value and graphically checked the phase with thelight curves from five evenings I got in a week. It's very easy to do such manipulation under an APL system. You are right, it would be very interesting to better consolidate our results. I see something similar in the variance of the reports madetothe CS database.In recent days you can see successive values like 3.33, 3.63, 3.3 3.5...for a given observer. It looks like an undersampling of such oscillation but it'sdifficult to combine the results of several observers as there are significant zero-point differences between them. I agree with your physical point, the light curve up-and-downs result from chunks of materialsgetting inand next leaving the stardisk. The "pause" after what looked like the third contact is somewhat a question-mark: is it a large chunkgetting in ? Or the disk had a thicker section we confused with the third contact ? CHARA did show the eclipsing disk did cover only the bottom part of the star disk, are we now seeing that bottom or theedgeon 3rd contact side? Allthis would need a lot moreconsideration and certainly the help of CHARA ! My data of the first half of the totality are not as good as the more recent ones and most of the timeI got only one series of image made within few minutesof eachobservation day. It seems to me difficult to extract a lot more than the OOE (well visible) from such data, next we had the poor period of the conjunction. Now I think the best is to continue observing as we have a couple of weeks remaining before getting in the short night and conjunction difficulties. Yours truly, Roger

thanks for telling more about your technique, Roger. Indeed, 200 pixels is a lot of defocus/spreading! That should overcome the effects I was seeing in my data. I think my problems defocusing were because I don't have the lens-to-focal-plane spacing exactly right, and the SLRlens was designed for a specific spacing. Also, I have the V filter, and a couple of CCDwindows, between the lens and focal plane, which the Canon designers did not contemplate. I plan to try adjusting the spacing -- I have bought some 0.002" shim stock. But I've also switched to a microlensed camera, which also addresses the problem. And yes, we say belt and braces, or belt and suspenders, in English too! Regards, Gary.

Hi Roger > My software includes an adaptive reading aperture and for > such evaluates the surface of the sensor that gets a > detectable signal from the star, the result is usually > something like 200 pixels. The star light profil is also > recorded for further check. 200 pixel in the star aperture sounds large, but this is equivalent to a star aperture radius of just 8 pixel = star aperture diameter of 16 pixel. But the pixelof smaller chip DSLRs and consumer digital camerasareusually pretty small. It might be also interesting to note that the defocused star images usually don't look the same inside and outside the best focus ! In both cases one one will get a 3D intensity profile that looks similar to a circular cylinder (refractors, glass camera lenses, ...) or a circular ring-cylinder (SCTs, Newtons, mirror camera lenses, ...). But the top surface might look different. In one case the brightness profile can look like a (positive) dome with the brightest point towards the center. In the other case the brightness profile can look like a (negative) dome with the brightest points at the edge. This will depend on the actually present optical aberrations of the scope and camera lens. The centroiding might be affected by this difference. Especially far off-axis as this will distort the brigtness profile. In the end one will usually prefere insidevs. outside the best focus depending on how the defocused star looks like in the corners where the aberrations are the worst. CS Wolfgang

Epsilon Aurigae was observed for 2.2 hours on April 26 UT (JD ~ 2455677.65) in u'-band and it was stable at the 2 mmag level. It's brightness was the same as on April 2, so maybe egress is complete.

Hi Bruce Good to see you back again. Its a cool u' TS on eps Aur. Your egress plot fits the one of Jeff's campaign well: http://www.hposoft.com/Plots09/UBBand.jpg Nevertheless there is a long gap between your last two measurements. During this time much could have happened regarding the egress and the OOE variations that can be in the 1-2 dmag range within one month in the U band.. Its well known that the disk eclipse of eps Aur is shorter as bluer the band is: http://www.hposoft.com/EAur09/EAurPredictions.html Its no surprise anymore that the 3rd contact was much earlier than expected. But it would be IMO still a real surprise for everybody ifthe gress in the U band would be just half of the expected and predicted value. As the Sloan u' and JohnsonU band filter transmissions arepractically identical, you could rereduce your u' band mesurements also to the Johnson U band just by using the Johnsohn U band mags: http://www.hposoft.com/EAur09/Starinfo.html I wouldn't wonder if you wouldn't need a more advanced color correction other then a simple constant offset for your reduced (andtransformed?)Sloan u' band measurements. It might not be "nice" to reduce the same measurements to two photometric systems, but its IMO OK in this case as the filter transmissions are pretty identical and one will be able to compar your SDSS u' band measurements to the Johnson U band measurements of other observers. Did /willyou submit your measurements to: CS:http://www.citizensky.org/submitdata/nonvisual or the AID: http://www.aavso.org/webobs and to Jeff's campaign: http://www.hposoft.com/Campaign09.html ? CS Wolfgang

Hi Wolfgang, If I understand well you are discussing the known issue of the PSF fitting some astronomers use. It's clear it's not a good idea when using photo-lenses ! In addition I use star trailing to increase the photo-electroncollectioncapacity,this is also a function of the defocus further than just avoiding the pixelisation issue. The technique I have implemented in my software is different, to make it simple, this is to determine the footprint of the star image on the sensor, by the way taking care of any aberration, trail, defocus... That footprint is a bit-mask thatis further processed to ensure it's compact, isolated bits due to noise peaksare eliminated and a guard is added. This ensure (hopefully) all photo-electronsof the corresponding pixels are counted and extra noise due to residual background pixels is avoided. Yours truly, Roger

Hi Roger. Your findings is very intresting. Do you have an epoch for maximum for the 0.0462 oscillations? I have data for the period with magnitudes varyingup and down, but only with sampling ofone per day. It would beintresting to seehow they correlate with yours. Thomas Karlsson

Roger, Nice work. Other rapid cadence work will be needed to confirm the reality of this effect, and whether it is limited to egress. In a 2009 proposal for MOST spacecraft monitoring, I estimated that

"A characteristic, minimum accretion event timescale can be characterized as the crossing time for material orbiting a B dwarf star, i.e. the stellar diameter (~2,800,000 km) divided by the 134 km/sec orbital speed. That timescale is ~21,000 seconds or about ¼ day, which is nicely placed in the MOST cadence “sweet spot” in frequency terms. Shot-noise like power is anticipated on both sides of this frequency, and detection of a power-law behavior on those timescales would be strong evidence for the accretion scenario. These timescales may be difficult for ground-based work due to airmass, seeing and diurnal effects.

p.s. Gary Billings published his study of this phenomenon in JAAVSO: http://adsabs.harvard.edu/abs/2010JAVSO..38..142B Abstract:Rapid cadence (every 42 seconds) photometry of ε Aur in September 2009, using a V-filtered 50 mm f /2.8 camera lens and an SBIG ST-7E camera, does not show short period brightness variations significantly different from those of the comparison star η Aurigae. On a scale of minutes to hours, the only variations detected are attributable to scintillation and differences in atmospheric extinction. (this paper was also presented at the 2010 Society for Astronomical Sciences Symposium).

Hi Thomas, The maximum found on three days 18/4 ,20/4 ,22/4 in the evening were at JD 5670.3619 witha period of 0.046183 (days shown in the inset of my slide).In anew set of series made on 25th evening the correlation was not as good but possible and the variance amplitude reduced to about 0.02 mag instead 0.04. Today I think that variance is welldue to the star system but not sure at all it's a well established oscillation regime. Considering the variance of closely surrounding stars, the instrumental SD of each, the SD of the interpolation error of the extinction gradientagainst their catalog values I do not see how such 0.04 mag variance could be explained by the scintillation or differential extinction. It would suppose some "flying shadow"has affected epsilon only, thatfor several days... Clear Skies ! Roger

Leaving aside the question whether there is a periodic high frequency signal, maybe it would be worthwhile to check first whether there is a higher variance now compared to other phases of the eclipse. Like this: take measurements from a single observer with many observations, and plot variance of measurements in a (say) 5 day window against time. Could that tell us something ?csHB

On second thought, to get an unbiased result, one would have to take the overall trend into consideration or else you will always get a higher variance during ingress and egress just because the values are spread out more. So a better alternative could be to fit the observations of the last (say) 5 days with a linear or quadratic fit and take the quality of the fit as our output to compare for different phases during the eclipseCSHB

I have had two nights with good conditions and took several series of images. On april 30 4 series with 15 min interval, each series 3x10 images with 2 min interval. On may 1 I got 6 series with 10 min interval, each series 3x10 images with 2 min interval. Attached is the result overlayed with the suspected oscillation. I must say that in my data I didn't find any significant correlation with the suspected oscillation. The individual spreading between the datapoints is relative high though compared to the expected amplitud, but not so big to completly hide any underlaying variation I think. I also analyzed the data for this two days togheter with the data from 8 previous days from april 17 and forward. For these days I only have 1 series/day. I use the the lightcurve and period analysis software Peranso for the analyze but I didn't find any big peak for any period in the interval 0.02-0.09 days. Thomas Karlsson

Thomas, thanks for your observations, I think it's now clear my oscillation hypothesis is not confirmed. Igot two acceptable observation sessions on 4/25 and4/30 evenings.Each series included 15 shots of 10.4 secondsand covered about 5 minutes.I makeone series every15 minutes starting as soon stars are visible in high background, the height of epsilon isthen about30 degrees only. It's clear we are now entering a difficult period for observation and this is probably the reason for the strong variance on a time-scale of one hour. Surprisingly the variance withina singleseries, in a five minutes time scale, remains not that high... On 4/25 I got 9 series. The correlation with a sine wave was unclear, two or three series were far from fitting. On 4/30 I got 8 series with no correlation at all. Anyhow the pause of the egress is now well established, some are even seeing ahint ofdimming ! I see the system about stable on a day by day basis (two hours average): - 4/18 3.353 (0.013) - 4/20 3.344 (0.015) - 4/22 3.348 (0.008) - 4/25 3.336 (0.008) - 4/30 3.347 (0.013) The pause is probably accentuated by the descending phase of the OOE. When would we see a restart ? Soon it will be very difficult to observe here in Burgondy... Yours truly, Roger