18 replies to this topic

[Awanderer]

Yesterday, we tested the Automatic Tilt Adjustment feature（BETA now） on-site for a Takahashi Epsilon 180ED user who had purchased the Wanderer ETA. This customer had been struggling with tilt issues for three years and had nearly given up. After configuring the parameters correctly, WandererETA resolved the problem in just 10 minutes.

 

The video record can be found here: https://www.youtube....v=6AdssOTdOhI

Don't miss this miraculous moment!

 

 

 

We used his imaging setup to demonstrate the impressive tilt-correcting capabilities of the WandererETA, working seamlessly with the NINA Hocus Focus plugin.

 

We first randomly adjust Wanderer ETA to make the camera severely tilted, as shown in the following two pictures:

 

 

In just ten minutes and three adjustment rounds, WandererETA delivered the following results：

 

 

The user shared this comment after the test:

I've been struggling to adjust the tilt of my Takahashi 180ED for the past three years, never quite getting it right. I tried several different focusers and even gave up at one point. But tonight, I finally solved the problem that had been bothering me for so long. At last, I can confidently place it in a remote observatory.

Edited by Awanderer, 13 May 2025 - 07:28 AM.

[deSitter]

What is the tilt problem? Mirror flop?

 

-drl

[dan_hm]

Impressive result. How does it work with HocusFocus if the adjuster has 3 points of adjustment rather than 4?

Also, I’d have to correct your claim at the end - the ASG electronic tilter came first.

[Dean J.]

What is the tilt problem? Mirror flop?

 

-drl

The primary mirror is fixed in the Epsilons.  The scope also has to be properly collimated.

Edited by Dean J., 13 May 2025 - 08:41 AM.

[Awanderer]

First of all, please note that this is automatic, not just electronic, and you don't need to decide how much you need to adjust which point. Even if you are talking about electronic tilt adjuster, WandererAstro released WandererMaster in 2023 with built in electronic tilt adjuster, although it is not sold globally. 

 

"How does it work with HocusFocus if the adjuster has 3 points of adjustment rather than 4"

It is calculated by WandererEmpire software, key of automation. In terms of precise adjustment, three point is much better than four point actually. Like you will never see a tripod with four legs.

Impressive result. How does it work with HocusFocus if the adjuster has 3 points of adjustment rather than 4?

Also, I’d have to correct your claim at the end - the ASG electronic tilter came first.

[andysea]

What sensor were you using?

[andysea]

Impressive result. How does it work with HocusFocus if the adjuster has 3 points of adjustment rather than 4?

Also, I’d have to correct your claim at the end - the ASG electronic tilter came first.

You can't define a plane with four points. You need three points, unless one is a extrapolated from the other three. I suppose that is why most systems use three points of adjustment.

[dan_hm]

You can't define a plane with four points. You need three points, unless one is a extrapolated from the other three. I suppose that is why most systems use three points of adjustment.

Octopi and ASG both use 4.

[Awanderer]

We understand the advantage of using 4 points. The four-point adjustment method feels more intuitive because an image naturally has four corners and also Hocus Focus gives the data of four points.

 

However, intuitive doesn't necessarily mean optimal. In fact, four-point adjustment is a compromise, designed for users who may not be able to precisely determine how much each point should be adjusted. However, now we are talking about automation, with WandererEmpire, our software leverages the sensor’s size, angle, and other parameters to build an accurate model, achieving an exceptionally precise effect in the video. Four-point will be an overconstraint problem, not good for calculation. I am not exaggerating, just law of physics.

 

Let’s explore why four-point adjustment, while intuitive, is fundamentally limited.

In a typical four-point setup—top left, top right, bottom left, and bottom right—you might think adjusting one corner gives you precise control. But in practice, you’ll encounter several complex scenarios:

Scenario 1: The top left and bottom left corners are firmly in contact, while the bottom right corner is suspended. Adjusting the top right in this situation causes the focal plane to rotate around the Y-axis—an unintended motion.

Scenario 2: The bottom left and bottom right corners are fixed, with the top left suspended. Adjusting the top right here results in rotation around the X-axis, again deviating from the intended movement.

Scenario 3: The top left and bottom right corners are fixed, and the bottom left is suspended. Now, adjusting the top right causes rotation along the diagonal line connecting the top left and bottom right corners. This is the desired type of adjustment.

However, even in scenario 3—the "ideal" case—there’s a catch. Unless your sensor (e.g., CMOS) is perfectly square, this diagonal-axis adjustment deviates from the theoretical ideal required for focal plane alignment. After modeling we have found this small deviation can make a big difference when tilting a fast telescope.

While it's theoretically possible to overcome these limitations using extremely high-precision engineering—such as synchronized control of four ultra-precise motors, a completely backlash-free mechanical design, and motors equipped with high-resolution encoders—this becomes exceptionally challenging. It’s particularly demanding when dealing with micron-level adjustments required for precise focal plane alignment.

Octopi and ASG both use 4.

[NebulaWAN]

A plane is defined by at least three vertices.

This is the case in 3D, so there are no drawbacks to manipulating a plane, regardless of the number of vertices, by three points.

For those who can't imagine it, try manipulating a plate with three fingers on one hand: it will be simpler than manipulating it with two fingers on each hand  !

The rest is just a matter of algorithms.

[Awanderer]

That’s right.

The same principle applies to a table or a refrigerator. A three-legged table is always stable because all three legs will rest evenly on any surface. But with a four-legged table, unless it is very expensive and well manufactured, you often need to add something like soft pads under the legs to keep it from wobbling.

A plane is defined by at least three vertices.

This is the case in 3D, so there are no drawbacks to manipulating a plane, regardless of the number of vertices, by three points.

For those who can't imagine it, try manipulating a plate with three fingers on one hand: it will be simpler than manipulating it with two fingers on each hand  !

The rest is just a matter of algorithms.

[Dan_I]

Interesting !  Tilt on a Newtonian like the epsilon is usually a combination of two things: collimation error (the focuser mechanical axis does not coincide with the optical axis) and mechanical tilt (the sensor plane is not orthogonal to the mechanical axis of the focuser). It seems that your algorithm attempts to correct both indiscriminately so if the collimation is not perfect you could end up with a tilted camera ?

[dan_hm]

We understand the advantage of using 4 points. The four-point adjustment method feels more intuitive because an image naturally has four corners and also Hocus Focus gives the data of four points.

However, intuitive doesn't necessarily mean optimal. In fact, four-point adjustment is a compromise, designed for users who may not be able to precisely determine how much each point should be adjusted. However, now we are talking about automation, with WandererEmpire, our software leverages the sensor’s size, angle, and other parameters to build an accurate model, achieving an exceptionally precise effect in the video. Four-point will be an overconstraint problem, not good for calculation. I am not exaggerating, just law of physics.

Let’s explore why four-point adjustment, while intuitive, is fundamentally limited.
In a typical four-point setup—top left, top right, bottom left, and bottom right—you might think adjusting one corner gives you precise control. But in practice, you’ll encounter several complex scenarios:
Scenario 1: The top left and bottom left corners are firmly in contact, while the bottom right corner is suspended. Adjusting the top right in this situation causes the focal plane to rotate around the Y-axis—an unintended motion.
Scenario 2: The bottom left and bottom right corners are fixed, with the top left suspended. Adjusting the top right here results in rotation around the X-axis, again deviating from the intended movement.
Scenario 3: The top left and bottom right corners are fixed, and the bottom left is suspended. Now, adjusting the top right causes rotation along the diagonal line connecting the top left and bottom right corners. This is the desired type of adjustment.
However, even in scenario 3—the "ideal" case—there’s a catch. Unless your sensor (e.g., CMOS) is perfectly square, this diagonal-axis adjustment deviates from the theoretical ideal required for focal plane alignment. After modeling we have found this small deviation can make a big difference when tilting a fast telescope.
While it's theoretically possible to overcome these limitations using extremely high-precision engineering—such as synchronized control of four ultra-precise motors, a completely backlash-free mechanical design, and motors equipped with high-resolution encoders—this becomes exceptionally challenging. It’s particularly demanding when dealing with micron-level adjustments required for precise focal plane alignment.

Well this is definitely interesting then. Do you expect to be able to ship units within the US market anytime soon? Also, what is the depth of the telescope-side M54 threads? I’d like to be able to screw in a ZWO filter drawer.

[guanzz]

Interesting !  Tilt on a Newtonian like the epsilon is usually a combination of two things: collimation error (the focuser mechanical axis does not coincide with the optical axis) and mechanical tilt (the sensor plane is not orthogonal to the mechanical axis of the focuser). It seems that your algorithm attempts to correct both indiscriminately so if the collimation is not perfect you could end up with a tilted camera ?

In fact, there is no axial error calibration with zero error in real life. Any tool will try to make the error as close to zero as possible, even the Catseye XLKP, which is the gold standard, so the final calibration is to fit the orthogonal relationship of the light cone reflected by the secondary mirror with the CMOS attitude. Of course, when the optical axis calibration deviation is large, the CMOS attitude adjustment cannot make it fully compensate for the orthogonality of the light cone, but this itself is out of the calibration scope of ETA. Any CMOS tilt adjuster cannot directly distinguish the two errors you mentioned. Therefore, the premise of using ETA or any other CMOS tilt calibrator is to use reliable tools to make the axial error as close to 0 as possible.

[andysea]

I still would like to know what sensor was used for this demonstration.

[Awanderer]

As is mentioned in the video, it was a QHY268M which is an APS-C format camera. We also tested a QHY600 full frame on a FSQ106 with success, we will post that video soon.

I still would like to know what sensor was used for this demonstration.

[Awanderer]

Already available from our US distributors. Although very limited stock recently.

Well this is definitely interesting then. Do you expect to be able to ship units within the US market anytime soon? Also, what is the depth of the telescope-side M54 threads? I’d like to be able to screw in a ZWO filter drawer.

[calypsob]

I think you should combine this with the rotator

[charles.tremblay.darveau]

This is really cool, following!