MSL - Astronomical Observations, Phobos/Deimos, planetary/celestial observations and more Options

[Gerald]

... you could integrate along the star trail to beat down noise...

Based on two Sol 635 MR Regulus images (one used to clean the other one by subtraction of the darkest of both), here a somewhat more detailed regression analysis:
This image is a 32x32 pixels crop of the cleaned version of 0635MR0027010010401805E01_XXXX.JPG, magnified (nearest neighbor) by 64, and annotated:


In a first analysis step each of the 32 columns of pixels is treated as a separate sample of data.
The green channel of each pixel is squared and taken as a weight to average the pixel position, and to calculate the standard deviation.
The results of this step are visualized as red or green lines (mean, mean + stddev, mean - stddev).
Lines are green, if the summed weight (using squared pixel values) are above a trigger value of 400 (maximum possible value is 32x255²), and the standard deviation is below 1.5 pixels.
Else lines are drawn in red, and discarded as invalid for further analysis.

The weighted means of the remaining 9 valid columns are interpreted as values of a function of the horizontal pixel position.
By a standard linear regression method, the equation of the best-fitting (minimal RMS error) linear function is determined.
The linear function is drawn in blue within the considered interval.

Here a diagram of the residuals, meaning the differences between the weighted means of the pixel columns and the linear regression function:


The standard deviation of the 9 residuals is about 0.157 pixels.
The standard deviation of the mean is determined by dividing the standard deviation of the data by the square root of the number of stochastically independent values.
Assuming this number to be about 4, regarding the standard deviation of the vertical columns being below 1.5 pixels for an interval of length 9 pixels, the standard deviation of the mean should be near 0.157/2 pixels = 0.08 pixels.

This approach can be improved for at least two reasons:
- The star trail is not horizontal, the applied method can be adjusted to the rotation.
- The used file is JPG compressed, and contains compression artifacts, which increase the standard deviation in comparison to lossless images.
Therefore further improvements even with a single image seem possible.

On this basis the parallax of Alpha Centauri appears to be close to the statistical detection limit of assumed 2 sigma of a single MR image.

[fredk]

Yeah, I would not be surprized if the statistical errors could get below what you need to measure even the N-S component of parallax. But systematic errors, ie the things that Deimos pointed out, such as pixel response, optical distortions, etc, can easily dominate over statistical errors. These systematics are presumably uncharacterized at the level you'd be interested in and may depend significantly on temperature. When you reimage the alpha Cen field at a later time, the rover would have moved and the position of the star in the sky would have changed, so the orientation of the stars in the fov would be different. So you'd need the optical distortions understood to extreme (sub-pixel) precision across the field.

Anyway, I agree with others that this is very fun to think about! Before going too far I'd suggest checking on the Martian ecliptic coordinates of alpha Cen, since the cross-trail (N-S) component of parallax may turn out to be very small. The next closest stars visible with mastcam are about twice as far away as alpha Cen (starting with Sirius).

[Gerald]

Statistical errors have been what I could check for with reasonable effort.
I share your concern with the systematic errors in the subpixel range. That's clearly beyond what I can do out-of-the-hip. With a series of Phobos images I've verified it down to about 0.25 pixels, below I got aliasing effects. I've some hope to do better with Junocam, but it takes months of development and testing.

Checking the "cross-trail component of parallax" would be feasible.

[paraisosdelsiste...]

Finally, Sol 956 sunset imges are arriving, and they are amazing:

[James Sorenson]

Here are my heavily cleaned up versions

M-34

M-100

[Ant103]

And my takes on it Finally a sunset imagery with Mastcam ! \o/

[eliBonora]

Hi, here's a short video
https://youtu.be/yJsukIk7TbA

[pitcapuozzo]

Are these the images of the Mercury transit? Does anyone see Mercury anywhere?

[mcaplinger]

Are these the images of the Mercury transit? Does anyone see Mercury anywhere?

I think the Sun was overexposed. It's hard to predict the brightness with this much airmass.

Even in the best of circumstances and with proper exposure Mercury is just about at the limit of resolution.

[paraisosdelsiste...]

Could this point be Mercury or just any other artifact?

[scalbers]

Could be a sunspot also judging from its size. This wouldn't be comparable to Earth's view of the sun though with Mars near the opposite side of the sun.

Wonderful sunset images a few posts up that show the blue Mie scattering associated with the dust size.

[Deimos]

I imagine that is 0956MR0042270020502228D01_DXXX. If so, it is a sunspot, hours before the transit started.

In the image in which the Sun is almost bisected by the horizon, Mercury would be geometrically (IIRC) right of center and just above the horizon. The area is saturated, so the <16% brightness drop is not detectable in that image. And that's the highest airmass of the bunch, with the shortest (1 msec) exposure.

[fredk]

As a reminder, have a look at the previous Mercury transit for how hard this observation is. And that time the sun was fairly high (over 35 degrees elevation) and there were many well exposed frames.

[neo56]

Curiosity imaged Phobos and Deimos between 1h30 and 2h on sol 964:


Here is an animation I made with the pics:

[Phil Stooke]

Two views of Phobos on sol 964. Each one is a composite of five frames. Stickney is at left in both, north roughly at the top. The south polar crater Hall is very clear at bottom in the right image, and visible in the other one.

Phil