From Newsgroup: alt.astronomy


I posted recently about seeing moving objects with TESS. While the
telescope was designed to spot variations in light that signalled the
crossing of a planet in front of its primary star(s), there is
obviously way more you can do with the images than that.

Almost right away by downloading sequences of images taken looking in
the same direction for hour after hour -- but due to bandwidth limits
usually only selecting images ~60m apart despite many more images
being theoretically available on the STSCI server -- and processing
them lightly (contrast and zoom dialled up to 111), you could see
interesting things. The first "movie" up on my screen showed a field
with a few stars each blown up so they were small white squares in
the image. In the first 10-20 frames of the movie first one, then the
2nd, then 3rd, 4th, 5th successively blinked for a frame, separated
maybe by a frame each time. This was a movie of something moving in
some random direction hour by hour over 10-20h moving slowly
across the field of view. Unusually, the direction of movement was not
parallel or even perpendicular to the ecliptic. The blinking came in
from the top left corner, moving toward the center of the image.
In the TESS telescope there is 45 deg either side of the main
camera sight made up of 4 smaller cameras laid out in a row, each
of those cameras made up from a 2x2 grid of 2kpx x 2kpx ccds (and each
of those ccds made of of a 2x2 grid of sub-sensors).

I asked some AI program to create a tool that would search through
that first movie and find all such occurrences. It found millions of
them in the ~100 images. It also found millions of other things going on. various ultra-dim flashes that popped up for a few frames then died down
and went back to black. The flashes were of all kinds of shapes
and sizes and durations. So we were totally spoiled for choice over
what to look at next. Cutting this part of the story short, the AIs
found lots of interesting things and had some ideas about what some of
them were and I posted a bit of that over the years and remains are probably still hanging around on one or other web server.

But now I'd like to go back and look at what we might be seeing.
Starting with the most obvious kinds of things, then maybe progressing
step by step to other things that may not be so obvious or -- let's be
frank -- likable. :)

I'll take as the first dataset for this journey the first movie I got
off the line. TESS sector 1, camera 2, ccd 3. No particular reason
but it came near the front of the tarball STSCI had online at the time. Sector 1 is still waiting to go online in the new web display but
sector 3 has been up for some time. It may or may not give you an idea
what that part of the sky looks like with a telescope with a 24 deg
vertical and 96 deg horizontal "field".

<http://kym.massbus.org/TESS/PANORAMA/s3/index3.html>.

(I noticed a few hoomins looking at the new display; you can press the
small images to get a slightly larger and a bit more hommin-oriented
image to pop up).

So the initial idea is to take one of those little squares covering
about 12x12 deg of sky, bit-pick it to get an average luminosity from
the 4M 24b pix and make a time series. The images are taken ~1h
apart for about 100h sometime in 2018.

Can we take this time-series and "see" anything moving in there? Does
the light variation for the whole ccd contain information about things
moving a lot closer than the nearest stars? E.g. asteroids, planets,
comets?

I expected the answer. I was shocked to find it worked. That is one
peachy instrument! It can apparently see almost anything in front of
it for at least 100AU out. :)

For anyone wanting to do their own number crunching here is the
sequence I have for 1.2.3:

y206192942 8202.54 y207015942 20112.4 y207092942 21937.2
y207155942 22266.4 y207172942 22177.9 y208012942 22842.1
y208032942 22900.5 y208212942 20111.3 y208215942 20177.3
y209052942 20720 y209065942 20586.6 y209105942 19392.2
y209212942 18013.1 y210165942 17475.1 y210215942 16572.7
y211002942 16949.3 y211015942 17351.4 y212005942 15478.7
y212042942 16042.5 y213045942 14588.8 y213095942 14060.1
y213205942 12116.3 y213212942 11937.4 y214055942 12989.3
y214082942 12590.5 y214105942 12192.3 y214135942 11999.1
y214155942 12328.9 y215042942 12150.4 y215055942 11982.5
y215162942 11367.8 y215232942 9712.83 y216045942 10342.5
y216065942 9938.43 y216105942 9147.41 y216122942 8935.22
y216132942 8861.95 y217092942 4787.84 y218002942 4474.35
y218042942 4494.14 y218092942 4481.5 y218102942 4482.22
y218152942 4480.1 y219022942 4488.77 y219032942 4484.52
y219102942 4488.22 y219235942 4499.38 y221035942 4438.84
y221045942 4437.32 y221195942 18062.7 y221205942 18425.2
y221232942 19483.1 y222005942 19959.6 y222022942 19915
y222052942 19706.3 y222065942 20049.7 y222115942 19509.5
y222175942 19337.7 y222212942 19353.1 y223002942 19976.8
y223022942 20059.3 y223105942 18149 y223112942 17933.4
y223135942 18162.2 y224092942 17192.2 y224115942 15919.1
y224135942 15982.6 y225042942 16210.9 y225085942 15908.4
y225222942 13376.6 y225235942 14148.3 y226025942 15094.9
y226072942 14684.1 y226095942 14136.8 y226155942 13897.5
y226202942 12158.2 y227182941 12779.4 y227185941 12487.3
y227225941 11940.6 y228122941 12519.8 y229002941 11536.8
y229195941 10993.9 y230012941 11711.4 y230075941 12790.9
y230102941 12556.4 y230122941 12016.2 y230145941 11528.8
y230215941 9438.11 y230222941 9392.22 y230232941 9421.15
y231002941 9546.21 y231032941 9858.26 y231085941 9414.35
y231095941 9187.78 y231215941 5272.95 y232092941 4993.44
y232132941 5008.69 y232172941 5037.32 y233035941 5132.58
y233142941 5357.21 y234025941 6268.09

The first number of each pair is the TESS date. The "y" I put here is
the year the image was taken -- 2018 in this case. Then 3d of
day-of-year, then hh, mm and ss. There is a big manual to describe
what kind of date this actually is -- it takes into account
distortion from gravity fields, light transit time from the centre of
the solar system, all other kinds of things. Just assume it is GMT
from here.

The 2nd of each pair is the 24b luminosity as calculated with a
histogram program -- i.e. something that calculates the colour of a
square and the count of the number of times it happens in the image --
subtract off a value assumed for "noise" (the individual TESS images
estimate noise coming in off each the the A, B, C, D sub-ccd areas of
each ccd and you do with that what you need to) then take the log of
that, divided by a number to make that go from 0 to 1, then make it
into a 16b integer.

As you can see from the sample images on the web-site, this can make an awful-looking image with blobs and wispy stuff all over the place.
Basically it magnifies the importance of the very very dim part of
the image -- the stuff between the stars -- and may also pick up dust
on the camera lens or the slightest reflection off the cowling or
chunks of astronaut piss or anything else drifting nearby within
100,000 km.

But now the fun part. Can we spot anything in that mess of numbers?
Or is the avg brightness of a 4Mpx image just too course to ever
capture anything? Well -- to borrow a phrase from a much older Capt
Kirk -- let's see if we can.

Striping the information out of each TESS image you can find out where
the whole camera is pointing (in terms of RA,DECL) as well as the
RA,DECL of certain reference points inside each CAM/CCD combination.
Put that in a database. You can then compare those positions with the
RA,DECL for every planet, asteroid, comet and space probe you also
might have in your database. If anything comes within 45 deg on a
given day or given hour of that day then there is a faint chance (you
might expect) that around that time one of the frames should show a
blip up if that object is supposed to be brighter than space, or maybe
blip down if it just happens to pass in front of something bright
previous visible in the camera in that direction.

But running the comparison for almost all significant objects draws a
blank. While the guys running TESS don't mind if they point their
machine at Alpha Centauri or Sirius for weeks at a time, they are very
very careful not to point it a Saturn, Mars, Neptune or Ceres along
with a whole list of other larger well-known things.

But when it comes to small asteroids and rocks, go to town! The next
step up was running tough time-series regressions between the
luminosity data (above) and the position of each asteroid relative to
the TESS boresight. We know the RA and DECL of the object relative
to Earth, TESS is near Earth, so we can just shove in that position
against the direction the camera is pointing in. A more careful
approach would bother to get the RA,DECL of objects as seen from TESS
circling in an orbit around Earth and moon. (The Horizons database is
fully capable of calculating all kinds of parameters from any
well-specified place anywhere in the solar system; but I don't have
the life expectancy needed to get all that downloaded to my home
machines; I selected Mauna Loa at the start and that's the viewpoint
for all my planetary data since ;).

So we run a WHOLE BUNCH of regressions and try to make them as careful
as possible. Since we're dealing with time series you have to take auto-correlation into account. If your X and Y both vary with time
there is a good chance any correlation between X and Y is just down
to that. Time-series regressions subtract out any correlation against
time from the Y before seeing if anything is left over for X to
influence. And there are many thrilling details related to trying to
match data up from different kinds of things accurate to the
hour. Turns out, usually, no 2 things ever are measured in the SAME
hour. You have to improvise. Read, interpolate.

But that all done we begin to find some asteroids ping TESS at the
time it was taking images of Sector 1. As the distance in
"great-circle distance" terms gets smaller and smaller -- as the
object moves in its orbit closer and closer to the main telescope
boresight location -- we find the avg lum (above) goes ping at some
point. Some asteroids look black and their position corresp with a
darkening of the lum data at the appropriate time, some are bright and correspond with a ping upward in avg lum at the right time.

Here are some results:

Asteroid X in R2
distino distino avlum 0.57394075
distpanopaea sdistpanopaea avlum 0.43546863
distpandora avlum distpandora 0.42655049
distastraea avlum distastraea 0.42210857
distbotolphia avlum distbotolphia 0.41601817
distdavida sdistdavida avlum 0.40291915
distfides avlum distfides 0.37532702
distirene avlum sdistirene 0.36997610
distarne avlum sdistarne 0.35899439
distelisabetha avlum sdistelisabetha 0.35542499

First cab off this rank is asteroid ino:

Initial IAU76/J2000 heliocentric ecliptic osculating elements (au, days, deg.):
EPOCH= 2456968.5 ! 2014-Nov-07.00 (TDB) Residual RMS= .24403
EC= .2059984946941728 QR= 2.178563016858798 TP= 2456778.3913759049
OM= 148.2723348304917 W= 228.0118065209791 IN= 14.21169424825384
Equivalent ICRF heliocentric cartesian coordinates (au, au/d):
X= 5.236801528210030E-01 Y= 2.291982088935166E+00 Z= 3.877520071629954E-01
VX=-1.089863934927695E-02 VY= 4.078410680544463E-03 VZ= 2.260620872680371E-03 Asteroid physical parameters (km, seconds, rotational period in hours):
GM= n.a. RAD= 62.9105 ROTPER= 6.15
H= 8. G= .010 B-V= .705
ALBEDO= .096 STYP= Xk

As the angular dist of ino's position rel to the boresight direction
deceased and/or increased there was an effect on the avglum. We can see
from the ALBEDO (thanks, JPL server and hoomins) it is quite a dark
horse. We might expect as it moves in the image might go
darker. Seems unlikely we could see any shine on that thing and the
only effect we'd expect is more shade if it steps in front of a bright
star. But you'd be wrong. :) The beta of the relevant model is +ve
and emphatically so:

beta = 7363.88 +- 2267.37 (90% CI)

T-test: H0: beta=0 against beta>0
calc T = 5.5; crit val at 90% 1.7 i.e. Pr(beta>0) = 99.9994
Rank test: H0: X and Y independent
calc Spearman = .90; crit val at 99% .48 i.e. Pr(order by X and Y same) >= 99%

So it seems just this little crude measure does see things and quite staggeringly emphatic about it the statistics are as well.

You can even setup experiments to see which "class" of asteroid
approaching the boresight direction gives off the best vibes. I
divided them up into major, minor, tiny, hi-incl, and hi-eccentricity.

The best result for each class looks like this:

Category best R2
highec 0.30444887
hiincl 0.30444887
tinyast 0.30444887
minast 0.22109282
majast 0.11584911

So it seems the high-eccentricity asteroids (>.1, just arbitrary
number) seem to be the ones that "most likely approach the camera
boresight" -- well you might expect that from the name alone -- and
the major asteroids are the least likely to make an impact because the
TESS folks like to steer the instrument away from areas that contain
Ceres et al.

So it seems we can see individual asteroids, even very tiny ones (my
data goes down to diam of around 40km for the smallest of the "tiny"
ones, arbitrarily designed that because they have a JPL catalog number
<1000). We can even average up all the positions of each class and
see if the AVERAGE position of small asteroids or major asteroids also
acts like a dim or bright obj and affects the telescope image. And,
yes, it does. The average position obviously entails the positions of
all the components so if the components affect the telescope image
brightness then the average must. The "average asteroid" is like a pseudo-asteroid. :)

Now what about other things. One "sanity check" I thew in was
fireballs. Both the AMS and NASA have catalogs of bright fireballs
seen by the military or astronomers both amateur and pro.

But it turns out the AMS database doesn't feature anything that made
the TESS twitch at all during the period it was watching Sector 1.
However, the NASA database did. Generally these feature high-energy
fireballs and are spotted by instruments or working astronomers or
the odd astronaut. They usually blow up high in the atm and NASA's
instruments like to note how bright the flash was and where did it
burn up.

And some of THESE in 2018 did blip the TESS image. But only a very
tiny blip. Unlike the models above, the interaction between NASA
fireballs and the avlum of the image had an R2 around 0.03. But the
stats again was quite emphatic the result was there. The indep
variable (the "X") was the log of the hour-by-hour rate of high-energy
meteors falling to earth as recorded by the NASA database. While we
don't know how the light got from the Earth's upper atm over into the
lens of TESS camera 1 we're 90% sure in at least 2 different ways that
it did. The beta was +ve. Meaning (the log of) more fireballs coming
in made the TESS image a teeny bit brighter. We talking an average
~20-30 fireballs coming in per year. So there were only 2-3 blips
during the period of the sector 1 images sampled. But all the blips
are in the right place, so who am I to argue with a TS regression
program?

In a later post we'll have a look at what else seems to be visible
just in our little crude "average brightness" data series.
Eventually we'll have a look at some moving new pictures. The old ones
are still under the old area of the web page but I've forgotten what
most of them are.

--
What if The Brain Doesn't Create Consciousness? Scientist Proposes It Might
Be The Fabric of Reality Itself
ScienceAlert, 10 Aug 2026 21:43

Finland is still rising out of the sea after the last ice age, so new land keeps appearing along its coast - but under Finnish law it does not automatically belong to whoever owns the shore in front of it
Space Daily, 10 Aug 2026 21:42

El Ni?o could bring highest sea levels ever recorded to the California coast Los Angeles Times, 10 Aug 2026 20:40
A strengthening El Ni?o could bring the highest sea levels ever recorded to
the California coast this year, raising the risk ...

UFO files: Watch appears to jump 25 minutes ahead after encounter
NewsNation, 10 Aug 2026 20:37
"These lights, along with the first-observed light, appeared to `sync up'
and travelled to the east/southeast," says the FBI ...

Behind the exit of DeepMind's CEO: low morale, a talent exodus, and
model delays
Fortune, 10 Aug 2026 19:36

AI agent hacks gym booking system while trying to get its user a spot
Android Authority, 10 Aug 2026 16:33

[Easy Fix!]
China Planted 78 Billion New Trees-and Seriously Messed Up Its Water Cycle Yahoo, 10 Aug 2026 01:08

SA launches royal commission into artificial intelligence
ABC News, 10 Aug 2026
South Australia is launching a royal commission into artificial intelligence, with the state's premier saying society is at a "fork in the road".

'Very little to like': Wall Street assesses surprise July jobs report as
stocks jump
Yahoo Finance, 9 Aug 2026 22:28

[Make The Bots Pay!]
Cloudflare says humans could become a "rounding error" as bots generate
1,000 times more internet traffic
TechSpot, 8 Aug 2026 21:31
[Any chance then that humans get internet for free?]

EU lays out $18B satellite constellation plan, with Spanish firm winning
key role
Breaking Defense, 8 Aug 2026 21:12

Perseverance Finds Organic Carbon Preserved Just Beneath the Martian Surface SciTechDaily, 7 Aug 2026 19:38

OpenAI Didn't Notice Its AI Agents Using a Message Board to Plan Their
Hacking Spree
WIRED, 6 Aug 2026 21:24
[F*K! This is how it allays begins!!]
--- Synchronet 3.22a-Linux NewsLink 1.2