# Why (the heck) is the basic angle of GAIA 106.5°?

This answer to Why does the Gaia space telescope have two main mirrors says:

According to the GAIA FAQs which does an excellent job: http://www.cosmos.esa.int/web/gaia/faqs:

Why is there an angle of 106.5 degrees between Gaia's 2 telescopes?

The choice of the so-called basic angle of GAIA was a non-trivial one. On the one hand, it should be of order 90 degrees to allow simultaneous measurements of stars separated by large angles on the sky. On the other hand, it should not be a harmonic ratio of a 360-degree circle (e.g., 60 deg, 90 deg, or 120 deg). Taking these considerations into account, acceptable ranges for the basic angle are 96.8 +/- 0.1 deg, 99.4 +/- 0.1 deg, 100.5 +/- 0.1 deg, 105.3 +/- 0.1 deg, 106.5 +/- 0.1 deg, 109.3 +/- 0.1 deg, 109.9 +/- 0.1 deg, etc. Accommodation aspects identified during industrial studies subsequently favoured 106.5 deg as the value finally adopted for Gaia.

I don't understand any of that.

Questions:

1. "On the one hand, it should be of order 90 degrees to allow simultaneous measurements of stars separated by large angles on the sky." Well any large angle is a large angle, why should it be 90?
2. "should not be a harmonic ratio of a 360-degree circle" I can imagine for instrumental reasons, but why are "acceptable ranges for the basic angle are 96.8 +/- 0.1 deg, 99.4 +/- 0.1 deg, 100.5 +/- 0.1 deg, 105.3 +/- 0.1 deg, 106.5 +/- 0.1 deg, 109.3 +/- 0.1 deg, 109.9 +/- 0.1 deg, etc." and what does "etc." mean here? Is there a pattern I'm not seeing?

Any angle with an uncertainty of +/- 0.1 deg can be reached with some rational number with a denominator of 1800 or less, that's as far as I've gotten.

• The down vote is for the question being too challenging? Too mathematical? Not every question can be "How many stars are there?"
– uhoh
Commented Apr 4, 2021 at 21:41
• My answer on April 1st would have been: "The bond angle of liquid water is 106.1 ± 1.8°, so using an angle inside the errorbars of this for GAIA sounded cool." That's obviously not the answer. I also do not see any pattern (yet) in the series of possible angle choice. Could it have something to do with the position of the maxima of spherical harmonics? Something like "we want want to compress the data of our measured view as spherical harmonics with highest possible resolution"? Commented Apr 6, 2021 at 14:39
• @B--rian speaking of the H2O bond angle: How to find the angle between Mickey Mouse's ears as specified in an official trademark?
– uhoh
Commented Apr 6, 2021 at 15:39
• @aliential the concern about harmonics and systematic errors may be related to heating from the sun causing mechanical deformation in the "optical bench". Since the rotational axis is inclined there is a constantly rotating thermal gradient across the huge structure, and since it has quite a non-uniform distribution of objects and reflectivities, this could result in higher harmonics in the thermal expansion/contraction and torques within the structure. It's quite an engineering wonder, there's a bit about that in this answer.
– uhoh
Commented Apr 8, 2021 at 15:14
• From what I can tell, it has basically nothing to do with details of the instrumentation. It's more to do with the mathematics of trying to convert an integer number of scan observations into (preliminary) coordinates along a great circle on the celestial sphere (the term of art seems to be "great circle reductions"). They had exactly the same issue with the Hipparcos satellite even though the satellite and the detectors were very different (the chosen angle in that case was 58 degrees, so there's evidently some dependence on details of the spacecraft, scan mode, etc.). Commented Apr 10, 2021 at 0:24

The choice of basic angle in Hipparcos and Gaia is related to the "rigidity" of the stellar reference system that can be constructed when connecting accurate positional measurements across widely separated parts of the sky. Consider the following steps: (1) The main reason to make measurements from space is to get above the perturbing effects of the Earth's atmosphere (there are also advantages in terms of gravitational flexure of the instrument, thermal stability, and seeing the whole sky - unlike an observatory on the ground). (2) But a single field of view is not enough to connect accurate measurements in one region of sky (say covering an area of around one square degree) to measurements made in another region. A rigidly connetected reference system across the whole sky is essential for determining absolute star parallaxes and proper motions. Therefore, Hipparcos and Gaia superimpose two fields of view, and make the measurements of two widely-separated fields at the same time. (3) What should the angular separation of these two fields be? A little thought will show that it should be quite a sizeable fraction of a great circle, say 60-90 degrees. Think of this analogy: imagine trying to create a map of the Earth with a 1-m ruler: local measurements might be tied together extremely well, but the distance from (say) Paris to Warsaw would be hopelessly inaccurate because tiny local errors just build up - it's the same when trying to measure a stellar reference system over the whole sky. (4) Having agreed that it should be "quite a large angle", are there any constraints on what this angle should be? Yes! Hipparcos and Gaia make their measurements essentially along a long sequence of precessing great circles. Consider measurements made over one great circle (360 degrees) spanning the whole sky. If the two fields were separated by 90 degrees, you would only ever connect FOUR regions over the sky in a single great circle (360/4). If the two fields were separated by 60 degrees, that would be a little better, but you would only ever connect SIX regions of the sky together (360/6). Extending this argument, you can imagine that it is much better to avoid small integer fractions of 360 degrees (1/2, 1/3, 1/4, 1/5, etc.) if we want to create a network of star observations around the great circle with good rigidity. By extension, so too are small integer multiples of these angles, such as 2/3, 2/5, 3/7, etc. of 360 degrees. More formally, angles to be avoided are 360 x m/n, for small integer values of m and n.

Detailed studies in the early 1980s formalised these sorts of arguments, resulting in graphs which show the "rigidity" of the stellar reference system as a function of the basic angle between the two viewing directions. Angles giving poorer great-circle rigidity (90 degrees, 60 degrees, etc) are separated by small 'valley' regions providing good rigidity. Any of these 'valleys' would have worked well for Hipparcos and Gaia, and the choice mainly rested on the physical accommodation of the optical hardware within the payload. Following detailed studies, we chose 58 degrees for Hipparcos, and 106.5 degrees for Gaia. Some more intricate subtleties aside, many other choices in the range 40-140 degrees would have worked equally well.

My weekly essays on Gaia scientific results are available at https://www.michaelperryman.co.uk/gaia-essays, and essay 172 (15 April 2024) on "The Basic Angle" says a little more on this. It includes a graph which illustrates this rigidity versus basic angle more clearly. Reading this answer while inspecting this graph should help your understanding.

• Thanks for sorting this out! Commented Apr 17 at 7:28
• It's great when a new user stops by and immediately addresses a long-standing question! After reading through, I thought to my self, "if only there was some kind of plot that supported "Some more intricate subtleties aside, many other choices in the range 40-140 degrees would have worked equally well." I was pleasantly surprised to see it right there on p2 of your lined essay. Thanks!
– uhoh
Commented Apr 17 at 9:42
• I wonder though, could you say a little more about what is meany by "rigidity". Is it the extent to which the covariances in position uncertainties are beaten down? # Commented Apr 17 at 10:22
• For these sorts of details, best is to refer to the two (hyperlinked) papers that I reference in the penultimate para of my essay (Hoyer et al 1981, Makarov 1998). Commented Apr 17 at 10:37
• Thanks, I'll look at those, but just to point out that the graph in your blog suggests that 106.5 degrees is an angle with LOW rigidity. So I suppose "good" rigidity here, actiually means low rigidity, which is not an intuitively obvious thing. Commented Apr 17 at 16:50

I did not find anything conclusive on the web, and I am yet waiting to receive feedback from ESA's contact page. In a German speaking IT forum there is something a bit less vague than the FAQ cited in the question:

The reason for the 106.5 degrees is not easy to see. It has to do with the fact that we are (also) doing an astrometric solution on a great circle and for this all angles 360x(n/m) [n, m = small integers] have to be avoided because otherwise one would have a congruence state that would hinder the solution. Values other than 106.5 degrees would also be possible. 90 degrees would be ideal for parallax determination, but unfortunately 90=360x(1/4).

This said, why not simply choosing 91° as $$\gcd(360,91)=1$$ (with $$\gcd$$ being the greatest common divisor)?

Maybe it is really only a numerology of the engineers, just like the Mars parachute Morse message. An argument for that would be that Gaia also has also exactly 106 CCD sensors. Or it is simply a logistical issue on how to mount things on the spacecraft. I am really curious what the official ESA statement will be.

• Okay so 106.5 is 360 times (71/240), so perhaps 71 is considered a "not small" integer. How about the others? This is a question about rational numbers and the largeness of integers, so an answer will be based on rational numbers and the largeness of integers as well. But you also want to be comfortably far away from a small integer rational number, the smaller it is, the farther you want to be away from it. 90 and 120 degrees are 1/4 and 1/3 and systematic errors those low harmonics may be strong. Not by coincidence our magic angle is almost equally far away from both of them.
– uhoh
Commented Apr 7, 2021 at 18:26
• @uhoh I guess you are asking about the other possible values in your question, right? Commented Apr 7, 2021 at 18:32
• "but why are 'acceptable ranges for the basic angle are 96.8 +/- 0.1 deg, 99.4 +/- 0.1 deg, 100.5 +/- 0.1 deg, 105.3 +/- 0.1 deg, 106.5 +/- 0.1 deg, 109.3 +/- 0.1 deg, 109.9 +/- 0.1 deg, etc.'"
– uhoh
Commented Apr 7, 2021 at 18:39
• The 0.1 deg error bars irritate me - it could be mechanical limitations which limit us to 0.1 deg steps. I am so curious what ESA will answer, if they answer. Commented Apr 7, 2021 at 18:52
• The focal plane detector is huge, of order 0.5 meters wide a focal length of 35 m the samples both star fields (of the two primaries) over a rotation of about .8 degrees, so not sure what the 0.1 degrees is from. Unrelated but just asked in Space SE: What causes GAIA rotational axis to precess the way it does? How exactly is this accomplished?
– uhoh
Commented Apr 7, 2021 at 21:25

EDIT- The Frame of GAIA is not symmetrical so it's not related to wave resonance. It can be related to the 360' rotations of the cameras:

1/ 90 degrees gives the most precise angle for triangulations of normal vectors and vector magnitude on a cartesian graph. You can try it out by programming 3D mesh, else by glancing sideways travelling on a bike: glancing at 90'C, you have the highest measurement of the movement of X relative to Y compared to looking at 45'. If you have binoculars on the bike helmet (i.e. a narrow field of vision) you will find that 90 degrees views gives the best relative difference.

2/ The second statement is incomplete, missing context: On the other hand, it should not be a harmonic ratio of a 360-degree circle (e.g., 60 deg, 90 deg, or 120 deg). Taking these considerations into account, acceptable ranges for the basic angle are 96.8 +/- 0.1 deg, 99.4 +/- 0.1 deg, 100.5 +/- 0.1 deg, 105.3 +/- 0.1 deg, 106.5 +/- 0.1 deg, 109.3 +/- 0.1 deg, 109.9 +/- 0.1 deg, etc. . The suggested values don't impart notable harmonic ratios in the same way that phi rotations give the best coverage of light for a plant:

• The CCD time is set to 4.4 seconds and can be changed to 4.47, rendering angular resonance NEARLY TRIVIAL.
• The rotation time is 1 degrees every minute, one rotation every 6 hours (4009 photographs at 4.4 seconds apiece), which can be changed to i.e. 6:20.
• GAIA has already rotated sideways by ~1 degrees every time the cameras have scanned round once, and the aperture of the CCD is 1.7 x 0.6 degrees, The resonance shifts mostly depend on those figures.
• The sweep has to also keep the solar panels towards the sun and the sun away from the CCDs and to provide a very constant thermal environment. some very decisive constraints compared to rotation resonance.
• The satellite frame that holds the equipment isn't regular or symmetrical, so any vibrations and movements that travel through the frame will have irregular vibration modes at angles other than 90'C

Here's a video of the complexity of the supposed "360 degrees".

https://www.gaia.ac.uk/science/parallax/scan

• I'm not so confident that all this "ESA is wrong" is really true. In Stack Exchange it's important to support assertions in answers using authoritative sources or at least going through the argument and the math in some details. Comparing GAIA to a plant isn't helpful, and writing "NEARLY TRIVIAL" and "IS A WRONG STATEMENT" in ALLCAPS is a little dramatic especially for an answer so thin on substance. This currently reads like an opinion and a little bit like a rant, can you elaborate on these points? Thanks!
– uhoh
Commented Apr 8, 2021 at 12:02
• This isn't correct. There certainly are small ranges of angles which are possible and have seen a plot showing the expected errors vs basic angle that shows the peaks and troughs and that the troughs occupy narrow ranges of angles. Commented Apr 8, 2021 at 12:05
• I don't think the CCD exposure time has anything to do with it. The important factors are the time interval between when a star appears on the detector and the length of time it is on the detector. Commented Apr 8, 2021 at 12:08
• The satellite rotates at most by 0.06 degrees for every image capture (4.4 seconds). Do you mean that it's important for the 106.5 angle or for the exposure time? What kind of resonance were you referring to? vibrations in the steel? I'm interpreting the suggested resonance issue as timing the camera wrong relative to position so as to miss some views. Commented Apr 8, 2021 at 13:06
• When you use the ' symbol do you mean degrees or minutes of an angle?
– Fred
Commented Mar 7, 2022 at 14:36