Motorized Telescope Mount (10″ Dobsonian)

I purchased an improperly-modified reflector telescope on Craigslist for a bargain. It was an excellent telescope; a 10″ mirror collects an enormous amount of light. But it needed work, and it needed a precision motorized platform to be able to capture long-exposure images of the night sky.

After fixing its previous issues and confirming that the mirrors were unharmed, I built the below platform, which was carefully-designed to rotate about an inclined axis that is parallel with the earth’s rotational axis.

The combined system, in working condition. A Bahtinov mask covers the aperture and is used for focusing.

10-15 minutes of motorized operation. The system has an angular tracking accuracy (mean value) of 1-2 arcseconds/minute. Sub-minute periodic error was never completely eliminated.

Tracking the Stars: “Poncet-” or Equatorial-Platforms

On a budget, it’s difficult to obtain both the benefits of a large-mirrored telescope (large aperture and light-gathering ability) and that of a smaller astrograph (ability to use smaller and more affordable motorized Newtonian mounts).

A common solution is to build a platform that rotates about an inclined axis. This is not an easy task, as nearly every dimension is calculated from a model, and many are critical. Each platform is built to match the latitude where the photos are being captured:

This schematic shows a platform built for a latitude of 58 degrees North.

The dimensions are chosen so that the axis of rotation is coincident or slightly below the CoG, so *very little* work is required to rotate the telescope.

CAD model showing the basic model for 27-degree North latitude. The inclination of the axis of rotation is chosen by varying both arc segment radii.
The telescope’s CoG was determined by estimation of the masses of the major components.

Early Progress

The previous owner had purchased a focuser that did not fit well, and hacked into the upper ring to install it. Their goal was apparently to shift the focal plane so that it was coplanar with the camera’s sensor. It may have helped, but not enough for the Canon 6D I intended to use. I took the focuser off, fixed the sheet metal, and sealed the gaps with permanent black neoprene adhesive foam from McMaster-Carr.

Distilled water and lab-grade sterile slide wipes were used to clean the primary and secondary mirrors.

The telescope, the day it was purchased, with no modifications.

Collimation adjustment knobs added, as well as spacers and springs to shift the focal plane to within “reach” of the camera’s sensor.

Modeling and printing the collimation screw knobs to fit standard hex-head bolts.

Constructing & Mechanizing the Platform

The platform build itself went through several revisions. I struggled with eliminating the numerous sources of error present in a machine intended to rotate a heavy object at rates that must vary less than 1-2 arcseconds per minute. This value describes the variance (error) in rate allowed, on top of the fixed 0.25 degree-per-minute rotation required to offset the earth’s rotation (known as the sidereal rate).

Early revision. Shows fork-and-pin coupling arrangement that drives the platform from a threaded rod.

Metal bands added to the arc-segments. This reduces rolling resistance against the 12 roller bearings.

First power-up. An Arduino Nano and stepper-driver board are used to rotate a threaded rod, which pushes a fork linearly across the base. The fork is coupled to the upper platform, “pushing” it about the inclined axis at the desired rate.

Attempting to diagnose the error in the image to the right. This early design had excessive radial runout, on the order of 25-30 thousandths (0.030″). This was likely the source of error type (B) to the right.

This image shows two kinds of error: A) tracking, or drift error, and B) periodic error, the sawtooth shape in the star trail

The early fork-and-threaded-rod system. The stepper motor was sourced from an old laser printer, and the drive electronics are an inexpensive Arduino Nano and stepper-driver board, soldered together on perfboard.

Early Drive Mechanism

The fork-and-pin system proved to have too many sources of error, but I stuck with the design for a while to eliminate as much remaining error as I could.

Working control box and wiring harness.

I machined the leadscrew using a Dremel tool mounted in the vise of my mill, and the leadscrew in place of an endmill. The goal was to reduce radial runout.

A junction box with holes for the stepper driver’s heatsink housed the electronics. A tinted Plexiglas window allowed me to view the battery’s voltage (regulated to 12.0v) and the steprate multiplier.

As I was not confident in this design being final, the inside of the box was very prototype-minded. Ease of modification was key.

An early image from this system. Stacked in the computational astrophotography program, PixInsight.

Improvements to Drive Mechanism: Threadless Ballscrews

For reasons I’d learn later, UNC threaded rod does not make a good leadscrew. I created a threadless ballscrew using 6 small ball bearings and an 8mm chromed linear rail. In this case, I viewed it as a superior solution: near-zero backlash, improved rigidity of the rotating rod, and an infinitely available set of pitches.

To adjust for inconsistencies in desired pitch and effective pitch, I simply added a potentiometer to “trim” the step-rate sent to the stepper driver. The analog value from the potentiometer was mapped to a range that was added or subtracted to the fixed, hard-coded steps/min value. I chose that particular value by calculating it from the apparent pitch, shown in photo #3. A thread pitch gage and light grease found that the apparent pitch was equivalent to approximately 22 TPI.

Besides (ideally) zero backlash, minimal wear, and minimal rotational resistance, threadless ballscrews have axial load capacities much higher than I expected.
This is seemingly due to both contact materials (OD race of the bearing, and the inner shaft) being hardened, resulting in extreme rolling-contact forces, and subsequently, high frictional forces.

Threadless ballscrews work by “clamping” a series of ball bearings about a central, smooth shaft. Each bearing is inclined at a slight angle, which results in linear motion when the inner rod is rotated.

Function testing of the ballnut.

Approximating the real pitch of the printed threadless ballnut.

Finalizing the Design

Countless modifications have been omitted from this project description. Many late nights were spent measuring run-out in thousandths of an inch with dial test indicators, taking test images to pinpoint periodic error, trimming step-rates, adjusting collimation, and more.

I eventually chose to keep the threadless ballscrew drive system. The ballnut was designed with a protruding threaded rod, which held a nylon roller. This roller pressed against a rigid stainless rod affixed to the upper platform, and drove it about the desired axis.

Time-lapse video of the threadless ballscrew traversing a length of the rail. Here, you can see the nylon roller pressing against the upper platform’s pin, driving it in an arc about the central inclined axis. The bearings that the upper platform ride upon are also visible.

Capturing images. I installed ‘Magic Lantern’ aftermarket firmware on my Canon 6D. This allows me to analyze the image data in histograms, bar plots, and includes a built-in intervalometer, which is in use here.

A final look at the completed machine. I purchased elastic fabric and sewed a skirt for the truss-tube section of the telescope. This was very effective in blocking stray light from spoiling images. A needle and graduations on the front arc-segment (+/- 10 degrees) show the safe travel range of the mechanism (about 80 minutes of data collection).

Computational Astrophotography

Stunning photos of celestial objects are possible with careful processing. While all of the photos below do not take advantage of this process, the deep-sky-objects and nebulae nearly require such processing. Essentially, a number of photographs are taken of the same object, and the data these images contain is integrated, summing the information stored in each coordinate to provide a higher depth of information.