...not sure if the lander mirrorball has been done before, but it is a great idea!
-pjam (Very much appreciating the discussion & problem solving in this forum!)

Beagle 2 had a wide angle that could be actuated over one of its cameras for exactly that purpose
http://www.mssl.ucl.ac.uk/planetary/missio...agle_DM_pic.php

The problem is - there's always a chance it could get stuck - thus rendering your camera useless.

That is the point of having a mirror ball. It acts as a ultra wideangle lens. And there is no additional optics or complex hardware involved. It is just sitting there much like the sundail on the rovers. It weighs close to nothing.

Can you give some visual reference for that direction on the panorama?

Sure. as soon as I get home tonight I can render a panorama with a white dot representiung the sun in there... (projected publishing time 21:00 CET)

perhaps a QTVR aswell. I think they read better...

Still a little fuzzy on the geometry, but seems like malmer's sun vector helps to narrow down the orientation of the lander. If the problem is to determine the lander orientation with respect to the comet's body-fixed frame, the sun vector gives a range of possible solutions. I.e., you could rotate the lander around the axis of the sun vector & get an equivalent sun-lander geometry even though the lander-comet geometry is changing. As malmer says, a similar sun fix from another panorama would pin it down, at least within measurement error.

On another subject--trying to correlate the comet's body-fixed frame to the 3D shapemodel, I tried to take an example of an OSIRIS wide angle camera shot and reconstructing the shot from SPICE information (which uses the model of a triaxial ellipsoid). Here's what I got for an image taken on Sep 12:

Click to view attachment

The reconstructed shot gratifyingly matches up well with the actual shot, with the equator shown as the boundary between red and white, and longitude quadrant lines drawn. I think that the red part is the northern hemisphere, less sure which end is 0 degrees longitude. Anyone have an idea?

Click to view attachment

here is an orthographic image looking from the sun direction...

That's almost art!

Thanks Malmer! I think I prefer that visualization to a panorama, since it induces less uncertainty.

Interesting that some images that should be in full sun illumination are in shadow.
So there must be something blocking them from this viewpoint.
Great work, at least in my mind the landing site is starting to make sense :-) .


A question: do we know at what time the panorama images were taken and over what period?

You have to remember that CIVA image 5 might be much closer to the lander than depicted in my image. I just let the images sit at an arbitrary choosen 1 meter from the lander. So the area is probably in shadow from philae.

The other images are either lit from behind or in shadow from the large wall...

How nice ! Thanks a lot Mattias

Nice! And that sun direction would seem to put to rest the early idea that Philae landed on its side. Of course the time of Rosetta's day when the images were taken must also be considered.

Agreed. For the images where the landing pads are in sight, we know that the surface is not closer.
For the rest, yes they could be closer, but that's where my attempt to position matching features and overlapping FOVs comes in. Those will give minimum distances to the lander.

Looks promising. What did you use to fix the sun direction - great circles (or equivalently planes in 3d) connecting shadows and casting objects? How many shadows? Which did you use? You could probably identify several with low precision from the high-resolution frame. That would give you a cloud of directions for the sun. The centroid of the cloud gives your best estimate for the sun's location, and its width an estimate of your uncertainty.

Did you use specular reflections from a leg? Do we know if the leg surface is smooth? Aren't there multiple possible reflecting surfaces oriented at different angles?

What are the heavy red, green, blue lines?

I found 3 shadows in the highresolution stitched panorama image. i then transfer those positions to the lowrez but straight "raw" frames. See green crosslike gizmos in attached image. (I use planes in 3d cause I'm a 3d kind of guy.) I could probably find quite a few more to narrow down the errors. This was just a proof of concept.

I rendered an image of a 3d landinggear model that I have built and the speculars end up in the same spot as on the real one. (the real ones are carbon fiber rods so they would be very smooth.) a cylinder is somewhat tricky in wow it reflects so im not using it to derive the position. i just used it to see if it "felt right"

the heavy RGB lines are the three planes that i use for my current solve seen edge on. (ideally they would be invisible from this viewpoint but there is a bit of measurment error as always.)

Click to view attachment

I saw a reference to imaging starting 5 minutes after landing. One thing to keep in mind is that the spacecraft had been rotated something like 35 degrees at a later time? I suppose we can still assume it is in about the same location.

Nice to see all these new image postings in just around 1/2 day's time.

If I'm not mistaken, they tried to get a second panorama but the site was in darkness. That was before the rotation.
The descent camera took (unreleased) images before and after the final rotation.

An interesting data to work in the future is the solar panel current.
As the Philae scientists mentioned, the current variations over time indicate shadows. And those can be used to reconstruct the surface around the lander ;-)

Compilation of positional clues from solar panels:

"During every 12-hour rotation of the comet, one of the lander’s solar panels is now exposed to an hour and 20 minutes of sunlight, while two other panels get the sun for less than 30 minutes each."
http://online.wsj.com/articles/philae-come...data-1415972442

"Yes, the come rotates. One day is about 12.4043 hours and one illumination period is about 4.5 hours where we landed. Panel two seems to be looking pretty directly at a nice comet afternoon and sunset, but we're surounded by rocks/cliffs that are casting some nasty shadows."
https://www.topiama.com/r/3298/we-are-worki...rol-and-science

"In Philae’s current orientation sunlight reaches the solar panels during a roughly 90-minute period each 12.4-hour “day.” For most of that time, less than 1 watt is available, but power output climbs to 3 or 4 watts for about 20 minutes."
http://spaceflightnow.com/2014/11/14/comet...ear-exhaustion/

"I work with the solar array and can say we have illuminaiton on the lid (facing "up") so we're at least not on our head! According to last info, we may be tilted towards solar array 1. We definatley saw no change due to APXS deployment, MUPUS deployment or SD2 drilling because the 4 day profiles we have data from (the last being well after SD2 drilling) are ridiculously identical. I can see shadows on the panels that are within minutes of each other each day (and our sampling rate is just over 2 minutes!). We have panel 2 completey free and see nothing in the CIVA picture, but there are shadows cast on it. On all other sides we see rock faces except underneath where there seems to be a very deep hole! So we were really lucky despite all the bad luck! The calibration on the CIVA cameras isn't exact so I can't give you exact numbers, but the walls look pretty close and they are definately all higher than the Lander! - VLL.
Oh, and we're probably facing NWish with panel 2 since we see the sunset (the sun rises in the East, sets in the West goes over North at noon). - VLL."
http://www.reddit.com/r/tabled/comments/2m...ht_control_and/

Also, at this Sun angle, there must also be something to the left and slightly behind us to cast this long, diagonal shadow (camera No.3)?

http://www.unmannedspaceflight.com/index.p...st&id=34384
http://www.unmannedspaceflight.com/index.p...st&id=34391

I think that the shadow you see are actually just the very blackness of space.

If you look at my QTVR it becomes slightly clearer..

http://mattias.malmer.nu/wp-content/upload...va-Panorama.mov

Here's a quick try at a panorama in Philae's reference frame with a cylindrical projection containing the X axis in the center. Y axis is 1/4 of the frame to the right and Z axis points to the top edge (zenith).

Click to view attachment

So as a sanity check to see what I may be missing, I think the XYZ sun vector from Mattias would place the sun near the Y axis (right of center) and halfway from the center to the top of the frame (altitude 46.5 degrees). However visually I would place the sun left of center at I've annotated in the image, above and just left of the boundary between cameras #1 and #2. There could be a discrepancy of sign with +/- Y axis.

ADMIN EDIT: Unnecessary quoting removed. Please refer to Forum rule 3.5.

I am not as clued up as you guys. Does this mean Philae is in a hole, a big hole ?

Maybe in a depression next to a cliff? To help get a better perspective it would be useful to rotate this image to an estimated orientation of the comet horizon (as was done earlier). This would put the sun more to the left of the ledge in image #1 instead of above. If we knew what part of the comet day this was taken (morning, noon, or afternoon) it would help me to get some perspective as well. I will try to rotate the image when I get some more cycles of time.

It look like the edge between light and darkness point straight to the Sun, and it suspiciously seems like the edge of something sliced with single plane. Less randomness than expected. Also, what is beyond the leg and under it? Shadow or empty space? If it is empty space, angle of that edge above is so great than we have some kind of overhang, like a cave, but almost nothing below lander feet. Probability seems greater if we assume that part of that wall is in the shadow of something on the left side.

http://www.unmannedspaceflight.com/index.p...st&id=34384

Updated my CIVA quicktime VR

link to QTVR:
CIVA with Sun


http://mattias.malmer.nu/2014/11/civa-sun-direction/

There is another useful clue. Camera 2 do not see ground at all. With 60 degrees vertical angle of view, and visual axis 15 degrees down, lower edge of the frame should be at -45 degrees with respect to the lander horizon. So, either the ground slope is greater, or visible part of the ground is in deep shadow. But deep shadow is unlikely. Consert antenna is fully lighted and Sun is falling directly on panel 2. We would see at least some signs of the ground in reflected light. Visible part of the ground should be very near the lander (45 degrees, equal to the height of the camera), and angle for light reflected from solar panel 2 is good. And anyway, Consert antenna is not so short for it to be fully lighted and nearby ground in complete darkness. Probability for the deep shadow with edge that must fall within the very narrow bounds on the ground is quite low.

Therefore, at least two legs are on the ground, but in the 60 horizontal degrees visible from camera 2, slope of the ground is greater than 45 degrees. In fact, several parallel bands of successively darker shades in the lower right part of camera 1 picture strongly suggest downward slope of some -70 degrees in the next part of the panorama.

http://www.unmannedspaceflight.com/index.p...st&id=34383
http://www.unmannedspaceflight.com/index.p...st&id=34382

Just to test a scenario (somewhat similar to Mattias earlier), here is a view assuming a tilt of 80 degrees so we see the comet horizon hopefully as horizontal. The highest tilt is with Camera 2 is pointing mostly to the sky. The sun by the illumination seems near the left edge of the image. If Camera 2 (near the antenna) is tilting up (right side of its field) and pointing S, then the sun would be about 45 degrees up in the northeast sky sometime during the comet mid-morning hours. This is slightly at odds with the statement that the solar panel #2 caught the sunset and was facing SW? Could depend on the missing portion and how far the cliff in the west extends to the northwest.

Click to view attachment

Center of the image may point SW. Perhaps these assumptions can be used to draw the diurnal arc of the sun in the sky, if we know the cometographic latitude. Are we near the equator?

Thanks for providing all those references in one place. Specially the general sunrise, noon, sunset sun motion and how it related to panels.

With that info, I converted illuminated hours to degrees, with one day being 360º. Of course, not accurate, but good enough to see how things fit.
And they do.

I marked the overall sunlight angles / strength / time at the site. I have no hard numbers, just interpretation of timings from the post above.

You have medium sunlight on panel 2, some sunlight on panels 1 and 3 and the strongest sunset illumination on panel2. You also have the overall site illumination.
I used the sunset position to align the angles:

Click to view attachment

So the pan seems to have been taken in the late morning. The panel illumination in the afternoon/sunset is consistent with the visible topography.

My interpretation is that indeed we are close to vertical, but on a steep slope. As mentioned above, there's no ground visible towards the west, so I'll speculate that we are "high" on a cliff.

According to SPICE, the first touchdown was in the morning and the second touchdown around high noonish, relative to the reference ellipsoid.

We also had a comment that the lander body's vertical axis appeared to be approximately perpendicular to the surface at the intended landing site, so communication with Rosetta was good. So if we were, say, 10 degrees from the intended landing site, we should be tilted 10 degrees from level at the final site.

Of course it's not clear how accurately these statements apply. Ie, 10 degrees? 40 degrees? That depends on the antenna pattern.

Another try based on some of the recent posts. Here Panel 1 is tilted up at 80 degrees. It might be consistent with panel 2 pointing NW and the sun is near high noon (invisible about 50-60 degrees up in the north - right of center)

Click to view attachment

Image centered about Northwest. Makes some sense with the higher cliffs in the NE blocking the morning sun. However these are jagged cliffs with some serrations going beyond the vertical if this projection is true. This looks less onerous in the projection shown 4 posts back.

I'd still go with a vertical lander. That's what ESA put in the released mosaic.
There's some tilt. From their illustration panels 1 and 2 are pointing a bit up. So I'd say about 10º from vertical.

We don't have hard data from the top panel, but I interpret the "illumination period is about 4.5 hours where we landed" to be related to it.
Meaning that the top panel received light for 4.5 hours.
If you translate that into angles, you get consistent results with shadows coming from the rock faces we see on the panorama, extending a bit above us.

To be clear on terminology, how is vertical defined here? The direction of the ~1 gram gravitational force? Direction to the center of mass? Surface normal to the reference ellipsoid? Lander with three feet flush against whatever surface could be defined as comet firma?

Vertical = gravity direction.
Imagine using a spirit or bubble level.
Vertical / horizontal is related to gravitational force. Even if the force is minimal, any particles would still fall "down" in that direction.
It's consistent with Philae keeping the landing gear down through the 3 landings and bounces. At least that how I see it.

The antennas are omnidirectional.
Applied to the Philae's attitude this shouldn't constrain too much. I'd say it tells, that the cone provided by the antenna (probably the upper hemisphere, with Philae's positive z-axis as upside), and the cone given by the surrounding rocks intersect similar to the the cone expected at the first touchdown site, at least with respect to the plane defined by Rosetta's orbit (resulting in an angle within this plane).
This should rule out excessive tilt like 90 degrees or more of Philae relative to the surface at the first td location.
Comparing communication windows with Rosetta's respective positions could help narrowing down the cone of Philae's visibility to S-band radiowaves, and constrain the local topography. It means at least, that the depression where Philae is sitting can't be too deep and too steep at the same time, quantitative estimates should be possible with SPICE and communication windows, as far as documented.

I can imagine the radio signal strength as kind of occultation pattern or slice, following a circle or ellipse centered on Philae.
Over different orbits, we may have slices cutting through the local topography at different angles.
If that data is being worked on, or if it will be made available is another question.

With a lot of water around there could be o lot of reflections also. Radiation pattern could become quite complicated, but it is certainly worth a try.

Short summary of positional clues, or at least how I see them:

*** From solar panels.

Available power depends on angle of the Sun and pattern of shadows. More power could be generated if the Sun is near lander “equatorial” plane. But that most likely mean that the Sun is in the lower position with respect to nearby terrain, so we have greater probability for shadows. Because of that, it is better to have more general summary, without mentioning any single solar panel, possibly something like this.

- For 4.5 hours, or 36% of the comet day, or 131 degrees of rotation, we have light on at lest some small part of some solar panel.
- Out of that time, we have somewhat less than 1 watt of power for 90 minutes, or 12% of the comet day, or 44 degrees of rotation.
- Out of original 4.5 hours, at some point in time we also have 3 or 4 watts for 20 minutes, or 2.7% of the comet day, or 10 degrees of rotation.

Another clue is shadows that change within minutes of each other, even on panel 2. With 2 minutes sampling rate, lets say he actually meant 4 minutes. Than, 2 degrees of comet rotation is enough to make a noticeable difference in power level, and that means we have, probably irregular, shadow edge on solar cell(s). That also imply something on the other side, away from the visible wall. However, it is not clear when, how often and how long these variations occur (and how large they are). I assume he would not mention it if it was not common and significant occurrence.

*** From panoramic images

- Lender is surrounded with wall on ~ 4*60=240 degrees. On 180 degrees wall seems to be very close. It is not clear how close it is on the remaining 60 degrees.
- On ~ 80 degrees we have empty space. In that angle, under the lander, there also seems to be mostly empty space. In another words, terrain is sloping “downward” with at lest 45 degrees angle, possibly even 70 degrees.
- Remaining ~ 40 horizontal degrees could also be empty space, or it could be wall in the shadow, or both.

Only clearly visible lander points of contact and support are one foot and Concert antenna.

We need time stamps of these pictures for the Sun position to be really useful. ESA should release them, it is most important missing fact now.

The "3 or 4 watts for 20 minutes, or 2.7% of the comet day, or 10 degrees of rotation" is mentioned to correspond to sunset and to panel 2.
Sunset is at W/ NW, so you can use that for the general orientation.
As I mentioned before, that also matches the open space in the general direction of panel 2.

Even by the numbers, it matches. From solar panels, illumination over 131º; from the images, clear space over ~80º+~40º~ = ~120º !

It'd be interesting to know the if top panel gets much sunlight, because there are two passive heaters there that can collect 12W of solar radiation. The electric heaters can produce up to about 5W. The "warm" compartment is insulated with two separate multilayer blankets, so maybe with the current heat input from passive and active heater the temperatures might not fall quite so low & the instruments could survive hibernation.

Here is the image in post #681 positioned so that North is in the center. The diurnal track of the sun from right to left is added, assuming Philae is on the equator. The sun's position should be at the top of the arc if it is noon. I can try various other assumptions as well beyond what is specified in that post. One thing may be questionable in that the ledge top at the sunset point looks implausibly close to the horizon. We should be seeing the background horizon there perhaps? If so this ledge should be at a higher altitude. Maybe the big cliff in the south doesn't quite go up to 74 degrees altitude? In other words, Philae's two feet should be closer in altitude to each other? This would suggest that Solar Panel #1 isn't the exact element pointing up and we should go at least some angle to the right of this.

Click to view attachment
High resolution version: http://laps.noaa.gov/albers/allsky/philae/composite.png

As per jmknapp's question it should be possible to use this diurnal arc to further model the solar radiation incident on each panel. My guess is the top panel faces NE, since the part of Philae tilted up looks to be in the SW. If that's true then sunlight would have to somewhat angle in past the cliff to illuminate it.

If the sun is north at noon then we are in the southern hemisphere. It was mentioned close to the equator, so I guess the sun will cross close to the zenith.

I believe things are different when we consider the sun's high declination of +37 degrees. The diurnal track stays in the northern sky the entire day, even if we are on the equator. Max solar altitude is 53 degrees. The nominal daylight duration is half the rotation period.

Here is a new iteration where a location 10 degrees right of the normal to solar panel #1 is the side pointing up the highest. This trends a bit towards Emily's diagram where the two legs of Philae off to the left were more at the same altitude. Still some mystery as to why the big cliff on the upper right isn't more directly lit. Perhaps the sun is really closer to it in the sky.

Click to view attachment
Full Resolution: http://laps.noaa.gov/albers/allsky/philae/...ite_m170a80.png

It would be reasonable to draw multiple sun arcs showing the seasonal change. The yellow arc gradually expands as we approach the equinox.

Tomorrow, Nov. 26th, 11 am reddit time (which is Eastern Time I think) or 17:00 CET, there will be a Reddit AmA (Ask me Anything) with the Philae and Instruments team from the DLR (German Aerospace Center) in the subreddit http://www.reddit.com/r/IAmA/

Here the announcement tweet from the DLR.

"The diurnal track stays in the northern sky the entire day, even if we are on the equator. Max solar altitude is 53 degrees."

scalbers, I stand corrected on that, thanks :-) I was suffering from geocentrism...
I have no reason other than "feeling" but you panorama orientation doesn't feel right.

I tried to match Malmer's sun location to your arch and it doesn't fit.
Also tried to adjust your arch over Malmer's pan and it works better.

Interesting to test all these assumptions. It's tricky with Philae tilted nearly 90 degrees. For example it seems that a great circle going from the sun to Philae'e XY plane between cameras 1 & 2 would hop over the zenith. So in that sense my latest sun location would be consistent with Malmer's XYZ coordinates after all, and also reasonably consistent with my sun position in Philae's reference frame (post #871).

QTVR with a couple of rays from the sun just to show light direction:

http://mattias.malmer.nu/wp-content/upload...norama_rays.mov


and also a non interactive panorama:

Click to view attachment

It is really not much more we can do unless we get some more hard data. If we had the time when the panorama was taken then we could start doing some things but as it is now it is just massively underconstrained.

Gave some depth cues to the CIVA image... I think we see further away than you would first think...

For me this really helps to narrow down the orientation...

Click to view attachment

Wonderful on the depth clues, it really does help orient things.

Wow, Malmer and Scalbers – fantastic work!

This might be a distraction from what you've been trying to do, but:
Malmer, your depth cues picture makes Philae's landing site look more like a real place than anything else I've seen.
Scalbers, your 1801x901 panorama is — by the same, non-scientific, criteria — my favorite rendering. It 'reads' like a decipherable landscape.

Would either of you have any interest in taking a crack at something like a quick and dirty integration of the two, in terms of tone and curvature/warping? That's something I'd LOVE to see.