|Proposer||(37223) Andre Kovacs (email@example.com) obscode: KADB|
|Assigned To||(3663) Dirk Terrell|
|Date Submitted||Oct. 7, 2022|
Dear AAVSONet Telescope Allocation Committee,
As part of the ExoClock project and also member of AAVSO, I would like to propose the observation of transits for the exoplanet TOI-216.02 ( using the Optical Craftsman 0.61m Telescope (OC61), located at the Mount John University Observatory (MJUO), via the AAVSONet.
The confirmed Neptune-like exoplanet TOI-216.02, discovered by the TESS Mission, is a particularly interesting target for follow-up observations due to the following factors:
In addition, the 24" OC61 telescope is particularly required for these observations due to the following factors:
Personally, I have already contributed to the NASA Exoplanet Watch/AAVSO Exoplanet Database with over 169 exoplanet transits (under observer code KADB) and also over 107 of the transits already published in the ExoClock database, using both the 24" robotic telescope at the Madrona Peak Observatory (MPO) and the 6" Cecillia telescope as part of the Harvard & Smithsonian MicroObservatory network, besides others from my personal observatory.
Finally, the observations of TOI-216 would require a high cadence, having the integration time adjusted accordingly in order to achieve a SNR~1000 (or ~1 mmag RMS), not to exceed an interval of 2 minutes between exposures and also having the minimum dead time between them as low as possible, in order not to impact the precision of the transit model fit to the data. Also, the specific observations should be scheduled according to the predicted transit times provided by the ExoClock schedule calculator service, and preferably done using a Johnson-Cousins Rc filter, in order to reduce systematics and minimize the strength of the limb-darkening effect, producing a transit light-curve with better defined ingress and egress transitions.
|Target||RA (H.HH)||Dec (D.DD)||Magnitude||Telescope||Observation Frequency||Expiration Date||Proprietary Term|
|TOI-216||4.932014||-63.26008||12.32–12.3||OC61||—||April 21, 2023||No|
Do you have an estimate of the transit depth of this target? If it is much less than 3-5 ppt, you will not see it reliably.
The transit depth is 4.06 mmag (~4 ppt) and the telescope is below the minimum aperture required for observation (19.11") according to the ExoClock project (that also considers the brightness of the star, the transit depth and transit duration, as described on Section 3.1.2 of "Kokori, Anastasia, et al. "ExoClock project: an open platform for monitoring the ephemerides of Ariel targets with contributions from the public." Experimental Astronomy (2021)"), as available here: https://www.exoclock.space/database/planets/TOI-216.02/
So, if the observing conditions are helpful, TOI-216.02 is expected to be observable using the MJUO, according to past observations from the ExoClock community.
Sorry, I meant to say that the telescope aperture is above the minimum aperture required for the observation (19.11") and the transit depth of 4.06 mmag (~4 ppt) is within the 3-5 ppt you mentioned.
Let's get some test images and see what things look like. 4 mmag will be a challenge.
Usually, a SNR~500 should give 2 ppt RMS for good observing conditions, with complete 1h pre-ingress and 1h post-egress baselines for systematics removal.
So, I would suggest we aim SNR>>500, preferably ~1000, so we would get ~1 ppt RMS when the conditions are good. But that all depends on the integration time required and avoid saturating the sensor, plus the overheads.
To be sure, we should simulate an outside of transit observation with ~1h of transit duration (plus the 2h of baseline) to check the RMS.
Regarding the actual observations, the transit duration is of 2.02h, and the session start times above are for the pre-ingress start time. So, the observations should have the following predicted start end end times:
- 2022-10-20 (24.9% Moon illumination at 99.7° distance) - from 08:14 (Alt: 30°) to 12:15 (Alt: 56°) UTC: Predicted 8% max counts increase during the session for R band;
- 2022-11-06 (96.5% Moon illumination at 80.6° distance) - from 13:06 (Alt: 67°) to 17:07 (Alt: 61°) UTC: Predicted 0% max counts increase during the session for R band;
- 2023-01-14 (56.3% Moon illumination at 99.6° distance) - from 08:35 (Alt: 67°) to 12:36 (Alt: 60°) UTC: Predicted 0% max counts increase during the session for R band;
- 2023-01-31 (78.5% Moon illumination at 88.9° distance) - from 13:28 (Alt: 47°) to 17:29 (Alt: 24°) UTC: Predicted 0% max counts increase during the session for R band.
The 8% of max counts increase for the first session must be accounted to avoid saturating the sensor.
Also, the seconds session might be particularly challenging with 96.5% Moon illumination.
Finally, the second and third sessions might not allow complete pre-ingress or post-egress baselines of 1h each.
Yes, test images will help us gauge where we are with this field. Remember that the noise considerations have to include the noise in the comparison star as well. If you have a suggestion for the comparison star, let us know. Looking at the field, this might be a little tricky.
Unlike the submissions to the AAVSO WebObs Database, the AAVSO Exoplanet and ExoClock databases do not require the use of a sequence of comps from AAVSO. So, ultimately, we could use the automatic comp star detection based on the brightness of the target from AIJv5, like usually done for unconfirmed exoplanets.
A 4 ppt transit depth may be doable? However, your logic concerning simply trying to achieve an SNR=1000 and thus being able to detect a 1 ppt transit is not born out be experimental evidence. Other factors make it virtually impossible to detect such a small transit depth from earth's surface. This is born out by SG1 team efforts.
An actual trial will tell? I assume you have taken into account the period of darkness in NZ for your recommended observation periods?
For instance, when using the Madrona Peak Observatory (MPO) 24" telescope, we have achieved RMS~2 ppt with SNR~500 easily (remember, RMS~1/SNR):
So, RMS~1 ppt with SNR~1000 is only more difficult, but not impossible. The minimum aperture calculator from the ExoClock team (calibrated from previous observations already in the database) shows empirically the same result, but with a more realistic benchmark.
By the way, the calculator from the ExoClock team is a bit conservative, sometimes the effective performance from the community/telescope is even higher than the minimum aperture computed, so the scheduler also suggests targets that would normally require an aperture even bigger, but with the disclaimer that the integration time must be at least 2x the overheads.
So, if the RMS is not as expected, I would suggest an investigation for unaccounted noise from the calibration frames or other systematics from the telescope+mount.
Things start to become impossible, from the ground using the current techonology, when we try to achieve sub-mmag precision, that is required for Earth-sized planets.
Regarding the times for the start and end of the observing session, the ExoClock scheduler takes care of selecting the transits visible during the night-time, based on the latitude and longitude of the observatory (-43.98557°, 170.46503°), much like what's done by the Swarthmore Transit Scheduler (https://astro.swarthmore.edu/transits/transits.cgi), but with updated O-C into consideration (something not present in the databases used by the Swarthmore).
The only thing that I could not account for was the horizon altitude at the site, but could be done, if provided.
PS: Remember, the transit times provided are UTC, not local time.
When do you think it would be possible to get some test images? At first, I would suggest a couple of exposures for different integration times (20s, 30s, 40s, 60s, 80s, and 100s) to check the SNR and estimate the appropriate integration time.
Based on the results from MPO, for this target having 11.903 R-mag, we should have SNR~1000 with 80s of integration time, or with 40s if we stack 2 exposures, considering 50% Moon, 1.5 arcsec seeing, 30s of overheads, and good atmospheric transparency.
Do you know what the overheads would be for the FLI ProLine PL09000?
What is the target's RA/Dec in sexigessimal nor decimal degrees?
I looked up RA/Dec.
Plan committed to OC61. You will get an SR filter, 60s test exposure soon. 4 hour time series scheduled as per transit date/times.
Sorry for the late reply.
I am only receiving e-mail notifications when the proposal changes its state (approved, allocated), not when new comments are added to it.
Here are the coordinates for TOI-216.02, in case you would want to confirm:
RA: 04:55:55.2548, DEC: -63:15:36.313
I received the 60s test exposures.
However, I noticed that they look very noisy, and achieved only SNR~360.
Also, AIJ is indicating that the histogram starts at -500 and ends at 10,102.5811. From previous experience, this might indicate some problem with the calibration process, since the histogram should start at 0 and end at 65,200.0000, for a 32-bit FITS file.
Would it be possible to retrieve the raw FITS images (20221025/AAVSO_P285_KADB_AUTO_ver3297/TOI216-02/TOI216-02_SR_202210/a0269.fits, 20221025/AAVSO_P285_KADB_AUTO_ver3297/TOI216-02/TOI216-02_SR_202210/a0270.fits, 20221025/AAVSO_P285_KADB_AUTO_ver3297/TOI216-02/TOI216-02_SR_202210/a0271.fits) and the FITS calibration images used (bias_temp-20_bin2_220914.fits, dark_30sec_temp-20_bin2_220914.fits, flatSR_bin2_220910.fits), for inspection?
Finally, the CCD readout mode used is 8 MHz digitization speed?
I was checking the FWHM obtained for the target and I noticed that the value at 3.83 arctic is much higher than what I was expecting for this telescope, considering a forecasted seeing of 0.86 arcsec for last night.
This makes me wonder if you are defocusing the exposures for photometry with this telescope?
This might help explain why the SNR is 5x lower than expected.
If so, I would recommend not to defocus for stars below 11 mag for high precision photometry of transiting exoplanets, particularly with a f/14 (slow) telescope. Would it be possible to observe focusing on the target?
If not, what would be the guiding error for this setup? Would it be possible to observe only with tracking and not guiding?
Ken would be the one to answer questions about how the observations are taken.
There is no defocusing used.
There is no auto-guiding used.
I can increase the exposure to increase SNR.
There is no issue with the calibration but I will try to get you the raw images?
fwhm of 3.83 PIXELS is larger than normal, but not by much. You need to remember that the pixel size is 0.549arcsec/pixel, so that 3.83 = 2.1arcsec. The images of yours that I looked at were in this fwhm range. I don't know where you got the 0.86arcsec seeing forecast; I've never seen better than 1.3-1.4arcsec at the telescope, and the typical is much more in the 2arcsec range.
There is no autoguider. The telescope tracks such that 4min exposures show no significant tracking errors.
A CCD has a bias offset in the signal. With OC61, that offset is in the 2000 count range, so a histogram of a bias frame will show a profile with peak near 2000 and then some typical shot-noise width. After bias subtracting and flat-fielding a science image, bad pixels can sometimes yield very negative numbers. We truncate the calibrated image so that any pixel less than -500 is set to the -500 floor. Similarly, if there is no bright star in the field, then you won't see a saturated pixel (65535 value). 10,102 seems reasonable to me for the peak pixel value.
The particular CCD used at OC61, the KAF09000, has significant residual bulk image. Because of this, a preflash LED sequence is done to fill the charge traps. This leads to higher effective read noise. So your signal/noise for short exposures (say, less than a minute or two) will not have as high signal/noise as you might expect.
I thought that we were using the low noise readout, though it takes significantly longer. I'll check and make sure that it is in use for the future. The difference is not important for exposures > 60sec.
We can give you the raw frames, but I'm not sure what you will gain from looking at them.
I usually use the Astronomy Seeing feature from Meteoblue. It usually works fine for all of my other locations.
Maybe, its forecast for Lake Tekapo, Canterbury, NZ, is off by 1 arcsec? It usually indicates from 0.8 to ~2 arcsec for this particular location.
I see, the significant residual bulk image might explain also some artifacts in the flats and science images of features that resemble shadows of nearby structures or reflections of light sources.
I expected to see some hot pixels from the sensor. That is why I asked about the 65,200.0000 max ADU counts.
According to the spec sheet for the PL09000, the reduction of the readout mode from 8 MHz to 1 MHz should decrease the read noise from 15e- to 10e-.
As long as we keep the overheads around 30s (currently is around 20s), I don't see any reason not to decrease the readout speed.
So, could we also have another try with exposures of 120s to check if the readout noise decreases?
Regarding the raw images, I was planning to actually check the gain, readout noise and dark current form those images, and see the computed values from AIJ for the calibration of the science images. That is why I asked for them.
Without those measurements, it find it difficult to judge from where the noise might be coming from.
Regarding the defocusing, the filter set used is par focal or do you refocus every time the filter is changed? I am taking this because I can see variations in size of what look like donut shaped stars between filters on the flats available here: https://occam.aavso.org/oc61/oc61_221025_images/flats.html
Actually the donuts are from dust on the camera window. I have not looked at the flats for quite a while (the image inspector is supposed to do so and inform us of issues), so it is obvious that cleaning is needed! I'll ask Nigel to take care of this.
The LED preflash is actually quite a significant noise hit. I bet you won't find any difference between 8Mpix/1Mpix readout rates. The new camera (QHY600) that I'll send out soon will be far better.
Mt. John typically has 1.5-2.5arcsec seeing, slightly worse than MPO61. If you ever find a 0.86arcsec image from any Mt. John telescope, let me know. :) I looked at Meteoblue, but didn't immediately find the seeing estimates. Could you send me a link?
There may be some hot pixels, but remember that we are binning 2x2 and so the contribution from a single hot pixel gets diminished in the sum.
The filter set is very close to parfocal, especially at the f/14 focus. We do allow refocusing between filters at other sites, but for OC61, all filters use the same focus.
I'll ask Cliff to put the raw science images and the bias/flat masters on the ftp site for you.
Ken, can you schedule some 120sec trial exposures for Andre?
The astronomical seeing forecast is hidden deep in their Outdoors and Sports section.
Here is the link for Lake Tekapo: https://www.meteoblue.com/en/weather/outdoorsports/seeing/lake-tekapo_new-zealand_2188526
I was searching around for references about the residual bulk image issue with the KAF-09000, and I found this paper that describes a possible solution via deep cooling: http://www.narrowbandimaging.com/incoming/033006_1_1314293318_1.pdf
I agree, probably binning 2x2 is tricking me to think that the histogram is wrong.
Please download the raw and calibration images involved from https://aavsonet.aavso.org/oc61/TOI216/. Just right click and save. Let me know if you need the calibration algorithm.
Exposure doubled to 120s.
Dear Cliff and Ken,
Thank you very much for sharing the raw images.
Looking at the raw frames you shared, I can clearly see the issues with the residual bulk image on the master dark and science images, as described in the paper I shared before.
Would it be possible to get a couple of bias, darks and flats (not the masters), in order to estimate the gain, readout noise and dark current?
I am afraid that I might be subestimating the dark current in my analysis, and impacting the SNR estimation.
The frame calibration from AIJ I managed to get similar SNR for the star, scaling down the 100s master dark. But, with the AIJ calibration, I can clearly see more sky background noise than from AAVSONet pipeline.
However, I could only get the same background level from AAVSONet pipeline if I enable bias subtraction (deBias) for the dark subtraction.
Could you please confirm if the master darks provided were debiased or not?
I was talking to Roland Casali, from the ExoClock project, about the issue with the KAF9000 sensor, and he suggested also the following paper for an alternative solution: https://www.techbriefs.com/component/content/article/tb/supplements/ptb/features/articles/10873
The procedure described in the paper is complicated, but it appears sound. The concept is that we need to test the pre-flash exposure time so that it does not significantly increase the dark noise at a given temperature.
What we might be able to do is use varying pre-flashes and measure the dark noise in the way it’s described. It may be that the extra noise we’re seeing is excessive pre-flash which then contributes with extra noise.
I think I found an explanation for the odd results using the RBI preflash, and a possible solution via software, in this document: https://pds-rings.seti.org/cassini/iss/iss_data_user_guide_160929.pdf
According to the author, the pre-exposure flood of the pixels, using the preflash and readout procedure, invalidates the usual way we do the darks calibration (averaging a set of darks and subtraction this average from the science images), requiring a more advanced approach for dark subtraction. As described at the bottom of page 68.
Instead, at the bottom of page 86, he discusses two alternative models for the dark noise (a model based on a sum of exponentials and a simple interpolation model), to be used for dark subtraction during the calibration.
Comments on this proposal are closed.