The atmosphere will certainly introduce the the largest errors (anomalous refraction, scintellation). Of course one has to take care that no direct sunlight falls onto the image sensors that these don't suffer thermal expansion and distortion.

If one has the option of using several telescopes, several smaller telescopes would probably better than a single larger one.

Ulrich, yes the atmosphere is certainly a major source of error. A follow-up star field image from the same location will be done about 6 months later to try and minimize atmospheric differences due to local topography. Two small refractors will be used, one 12.5 cm (f12.5)and the other 10.8 cm (f15.6). These will be attached to the same mount. Fast sampling (relatively to photographic plates) will be used to minimize smearing due to scintillation. Typical photographic plate exposure was 10 seconds+, my exposures will be 1 second or less. Ideally one would want to take images at a rate 20 fps or more to allow 'lucky imaging', I still need to experiment to see what the maximum rate should be. It will have to be a trade off between star field density and image capture rate. Thermal expansion as you mention is a problem as it could also change telescope focus if the tube lengthens due to expansion. A very accurate automatic focussing system is therefore required.

Do you have the time and opportunity to test the stability of relative star positions with your telescopes and cameras by night-observations at home? Since gravity should shift star positions only in radial direction any anomalous refraction and thermal lensing should shine up as lateral shifts. An algorithm could filter out the radial shifts and match them to a theoretical model obtained from GTR. Interesting project. Good luck!

Yes, I will make extensive tests beforehand to characterize the system. The deflections will be between 0.8 and 0.6 arcseconds for 2R to 3R. One can calculate the refraction using standard models but would have to build in some type of calibration by using reference stars further away from the Sun. Thanks for your comments.

I think that the most sensible thing would be to try and estimate higher order terms by comparing optical and VLBI estimates for \gamma (i.e., to estimate the " gradient of spacetime curvature" between 2.5 R_Sun and the 10+ R_Sun region that dominates the VLBI data). If your numbers are correct, and if the atmospheric seeing varies on a time scale < 1 second, you might be able to get a 10^-3 estimate of gamma, which would be interested compared to the VLBI (although not really competitive with the claimed errors from Cassini Doppler test, which _is_ based on data very close to the Sun).

Hi Marshall, last time I chatted to you was at USNO! Yes, estimating Gamma will be part of the exercise. I have done some estimates of Gamma and Beta using SLR data, and the results were also a bit weaker than VLBI; they are at the7x 10^-4 and 9x10^-4 level for LAGEOS 1 and 2 respectively. Will have to see what is possible with the optical tests. It is essential to capture a larger number of stars than was possible with photographic plates to get a good gradient estimate. The ultimate would be if one could detect the Sun's J2 spherical harmonic coefficient (oblateness) in the spacetime curvature gradient data. It must be embedded, but may be too weak to extract. The Cassini Doppler test has very good results, but I am not sure about the realism in the errors. VLBI is very robust so I have more faith in the VLBI estimates.

For the sake of interest, I add a link to a table and plot I made for a chapter in Sciences of Geodey II, of VLBI estimates of Gamma, Marshall's 1997 results are included. There has been a remarkable improvement during the last 30++ years. Just by chance, the plot starts about when the last optical measurements of the light bending around the Sun were made. The VLBI show a factor of ~400 improvement...the question is will a new optical approach show similar improvement...even a factor of 10 would be nice!

http://geodesy.hartrao.ac.za/site/en/resources/72-gtr-info.html

As there may be some interest in this project, i.e. testing the General Theory of Relativity, I thought it might be a good idea to keep a running commentary here on the development of the telescope system etc., maybe those who are interested can help with some advice on the way. So, I add a picture here of the 125 mm refractor as it is now. It is not a new telescope, it was designed and built by Mr Bernd Spaeker, of Borken, Germany. I have all the designs and receipts, as he kept careful record of the project. The telescope was named Astrobot I. The first year that the receipts start is 1973. I bought the telescope a year ago. It was not complete and I had to remove the drive section and power supplies as well as the control box. It could originally be steered in RA and DEC and could track in RA at sidereal rate. All of the original electronics were removed.

The telescope functions (power supplies switch on, focussing etc.) are now controlled via a microprocessor (ATMEGA2560) and steering is done using 2 Ingenia Venus servo controllers. I will add some pictures of the control sections.

In the picture of my previous post, it can be seen that the telescope has been raised, this is done via a jack-screw that can be locked. This allows the height to be adjusted, in practice I will use this to reduce the packed away size of the mount. So, I have bought a piece of mild steel hollow bar, which we have machined to have a flange at the top and bottom. This flange will be split in 2, be inserted into the jacked-up mount section and will be very solid and stable. The inserted picture is the of the steel insert being turned in the lathe to its correct diameter.

It would be necessary to investigate the effect of rapid temperature changes on your setup. The graph shows that the temperature drops during the eclipse at 5 degrees for one hour . I would suggest to put the big foil screen to shade your installation, which can be easily removed when starting observations. Do not forget that it is an island of Spitsbergen , and due to the drop in temperature the fog can sit during the full phase. Elena A. Makarova who participated in many expeditions to observe solar eclipses has told me about that. For observations on the island the higher place is favourable.

Hi Igor, good idea! Yes, I am planning to place the telescope in a tent type shelter, with an opening to the south. Having a reflective foil to effectively shade the telescope is a good idea.There is a high rise to the south, and I may be able to access this, but I am not sure whether it is practical. According to those who go there regularly, it can either be total sunshine or totally overcast that time of the year.

The sudden temperature drop can also cause dew to form on the optics, so I have to ensure that the telescope has stabilized thermally as far as possible.

I include a photo to the south, taken from the website kingsbay.no. It is a very beautiful spot on the Earth. If I can somehow access the mountain, I may escape low lying fog, on the other hand, I may get ridge cloud. Fortunately I plan to be there 3 weeks before the eclipse, so that I can scout the area very carefully. I will make the telescope assembly independent of mains power, so that I can locate the set-up anywhere.

The flanged section of the telescope pillar has been turned to size on the lathe and the inner diameter cut to size. The 12 (6 top, 6 bottom) bolt holes will now be drilled to 13 mm diameter, the stainless steel bolts are 12 mm Allen bolts. Wall thickness of the flange is 9 mm. It is very strong and rigid.

Today we are painting the new flanged section, the milled section has been cut in two and drilled. I will fit it later today, as a small rubber strip needs to be glued to one section, which will function as a seal of the gap created by the cutting of the tubular section.

Just as a side comment / question. There are a number of solar observing satellites in various orbits some of which have coronagraphs. Is it possible there is a great deal of such data available that hasn't been used for this purpose? Removing the earths atmosphere and having a permanent eclipse has got to reduce errors considerably. Heres a link to one such device http://lasco-www.nrl.navy.mil/lasco.html

Hi Paul, I checked that site, very pretty pictures! The problem is that these pictures are once more of the Sun and corona, just as the old light bending tests were when done with photographic plates, the contrast is very weak so the stars are not visible except maybe bright ones. The GTR experiment also needs to be very carefully controlled, with reference star fields taken form the same place using the same instrument. The test I am planning will only take images between 2R and 3R, to try and reduce the effect of the corona and to capture a large (relatively) number of stars.

Thanks for the comment!

The flanged pillar section was installed and bolted down this morning (31 Jan 2014).

It fitted perfectly...

The line in the centre is a thin rubber strip to keep out any dust. Now the tube clears the box. I pushed and shoved the telescope by hand a bit, the mount is rock solid and there is no indication that the lengthening of the pillar has introduced any flexure.

One of the sources of error is focussing. At the moment I have a MEGA2560 controlling the focussing motor, which moves the rack and pinion focussing mechanism. This can be setup via my control interface (written in C using Microsoft Foundation Classes, yes MFC). One still needs someway to do the focussing as accurately as possible, so I am investigating some focussing masks. Here is a simulation of a Bahtinov mask for the telescope using Niels Noordhoek's software "Maskulator".

A few days ago, while walking with my dog, I made the experiment of trying to think your thoughts, Ludwig, and ended up horrified. My refractor has a resolution of about 1 arc second, so I have to achieve subpixel measuring accuracy somehow. My refractor has a plane field of view of one solar radius at best, so that I have to patch many exposures together to cover the 3R field under consideration. How shall I produce reliable results during the short time of total eclipse under these circumstances. A terrible task indeed.

Hi Ulrich, yes, the atmospheric seeing conditions will probably not even be 1 arc-second, so the atmosphere will limit resolution. So, the idea is to use a camera with small pixels (1.55 um) and so have a 0.2 arcsec 'resolution' per pixel. As nothing will be resolved in reality at this level, even if my telescope was much larger, I will have to rely on the light energy distribution on the pixels, and then detect the difference in pixel captured energy in the deflected and non-deflected star image.

You are right in the patching together, it will be non-trivial. So in order to do some 'geo-referencing' , two telescopes will be used. One with a wider field of view so as to capture some brighter stars, these images will then be used to geo-reference the smaller fov images.

Concerning the reliability of the results, well, I just don't know. There are many variables, many problems, and in the end it might just be cloudy with zero returns. One can of course do some tests beforehand, and evaluate this strategy, perhaps during full Moon as this is about the brightness of the sky during the total solar eclipse. It is a pity it is not a longer eclipse, that the Sun will not be at a higher elevation etc. So conditions are not ideal...I think they were ideal in 1919. I am going to try and get access to the hill towards the south, it is ~475 m higher. There is an atmospheric research station there, access is by cable car.

I attach an image here, this is basically the 13.99' x 7.38' fov, which the 125 mm refractor will have, if you moved the telescope to the Sun's edge.

This is the 50.62' x 33.77' fov as seen with the 108 mm refractor. Angular size of 1 pixel ~ 1 arc-sec.

Hi Ulrich, take the dog for another walk, there is also an interesting effect to consider, which Sergei Kopeikin mentioned to me. " By the way, there is a very interesting relativistic effect associated with the comparison of the aberration of light and the aberration of gravity in the solar type gravitational bending experiments. The aberration of light displaces the visible position of the sun in the sky. The question is - will gravitational deflection of light be sensitive to the present position of the sun as it follows from the Newtonian equations of motion or do we have to take into account that gravity does not act instantaneously and the position of the sun, as measured in the gravitational deflection of light experiment, is also shifted due to the aberration of gravity."

Hi Ludwig, since aberration of light shifts both the spherical positions of stars and of the sun (for the sun, astronomers would use the formula for 'planetary aberration' but the difference to 'stellar aberration' is tiny) the gravitative shift pattern of stars will be radially symmetric to the center of the sun in camera observation. This is what I think is plausible and is, as far as I remember, the result of more serious thoughts of mine many years ago. Not impossible that on the next walk my dog will tell me that I'm wrong. Thanks for the interesting message!

Last night I submitted a proposal to the National Research Foundation (NRF) of South Africa for some funding for this project, mainly to cover freight and travel to Ny Alesund.

I have bought 2 cameras, one CCD (57% QE) and the other CMOS (72% QE) during a recent visit to France (visited Observatoire de Paris and OCA). Field of view of the CCD is X: 0° 39' 34" and Y: 0° 29' 47". pixel size 5.4 micrometre, will be attached to the 125 mm refractor, and the CMOS is X: 0° 10' 18" and Y: 0° 7' 38", pixel size 3.75. This then gives a resolution of X: 0.7”, Y: 0.7” on the 125 mm refractor (pixel per FWHM=1.55 ) and X: 0.5”, Y: 0.5” on the 108 mm (pixel per FWHM=2.64).

Estimated accuracy for differential astrometry is ~ 163 milli-arcsecond for 3 sec exposure on the 125 mm refractor for stellar separation of 200 arcsec (i.e. working in the isokinetic region).

Current tracking precision on the mount with my software is ~0.06 arcsec standard deviation.

Some further progress on the telescope, a 10.8 cm (4.25 inch)  (f14.8) has been added to the mount. A wedge has been fabricated which is adjustable, this can be set to the observer's latitude via a worm gear. The 10.8 cm refractor can be adjusted to be aligned with the 12.5 cm (5 inch). I am still working on the tracking software. Within a month or so I plan to do some tests using the two cameras. The mount easily handles the two telescopes, but of course the actual tracking accuracy will depend on the combination of software, gears, servo controllers and mechanical characteristics of the mount/telescope combination.

This project at Ny Alesund has been cancelled due to failure of my grant application. So, I  am planning to conduct the experiment in the USA during the August 21, 2017 total solar eclipse.

Interesting..

Ludwig Combrinck .. Please share me the best answer might you trust...

Regards…