All Posts
· 13 min read

Designing Telescopes for 3D Printing

There’s a moment in astrophotography where the first sub-exposure comes in and you see structure in something 1,500 light-years away, captured through an instrument you designed on your laptop and printed in your office. A $60 Chinese parabolic mirror, some carbon fiber tubes, and a few hundred grams of PET-CF filament. The Horsehead Nebula doesn’t care what your telescope cost.

I’ve published two astrograph designs on MakerWorld: a compact 114mm f/3.95 I call the Squat, and a larger 150mm f/5 called the Big Guy. Both are Newtonian reflectors designed specifically for imaging, not visual observation. Different constraints, different design decisions than a visual scope.

Getting here was a process.

How this started

The first time I saw the Milky Way with my bare eyes was in the Arizona desert. I was on a road trip across the western US with my wife and a couple friends, driving this shitbox RV from state to state. Colorado, Utah, Nevada, Arizona. We stopped somewhere with Bortle 1 skies, true darkness, and I looked up and there it was. The whole band of the galaxy, visible. I took some pictures with my iPhone and something clicked. I didn’t know it yet, but I was hooked.

A couple years later I picked up 3D printing as a hobby. A friend showed me a cheap 114mm visual telescope you could print for about $100 in materials. I built it, looked through it, and immediately shifted into “I wish I could take pictures through this.”

That led me to the OpenAstroTracker, an open-source 3D printed star tracker. You mount a camera on it, it compensates for Earth’s rotation, and you can take long exposures of the night sky. It worked. Sort of. Random failures, camera too heavy for the printed gears, limited to about 10-second subs before tracking errors smeared the stars. Enough to get a taste of what was possible, not enough to do it well.

I looked at the price of proper mounts and telescopes and scoffed. Thought I could do it myself. That’s the moment the rabbit hole opened. I started designing my own astrographs, iterating through failures, learning optics and mechanical engineering by getting things wrong and figuring out why.

Eventually I shelled out good money for a proper mount (you can’t 3D print sub-arcsecond tracking, I tried), but I was happy with the telescopes I could design and build. The scope is where the creative engineering lives. The mount is where precision matters too much to DIY.

Mirrors and the price-to-performance question

With DIY telescope building (the hobby calls it ATM, “amateur telescope making”), the mirror is everything. It’s the only precision optical element in a Newtonian. Everything else, the tube, the focuser, the spider, the baffles, exists to hold that mirror in the right place and keep stray light away from it.

The question isn’t “what’s the best mirror?” It’s “is it good enough quality to produce usable images, and what does it cost?”

For the 114mm, I use cheap Chinese parabolic mirrors off AliExpress. d114f450 mirrors, usually around $60. These are surprisingly good. Not professional-grade optics, but the results speak for themselves. The images on my site were taken through one of these mirrors. Orion Nebula, Horsehead, Rosette, all with a $60 primary.

For the 150mm, I stepped up to a GSO mirror. Once you push above 150mm aperture, I’ve found GSO to be a good baseline. The Chinese mirrors at that size can still work, but GSO’s newer mirrors hit a sweet spot on price to performance. You’re paying more, but you’re getting consistent quality you can trust for longer imaging sessions.

The focal ratio matters too. The 114mm at f/3.95 is fast, which means it collects light quickly but is less forgiving of focus errors. Tolerance scales with f-squared, so an f/4 system is four times less tolerant than an f/8. The 150mm at f/5 is more forgiving, which matters when your structure is 3D printed and thermal expansion is a real variable.

You can’t just put a mirror in a tube

When I started designing my own scopes, I thought the engineering was straightforward. Tube. Mirror at the bottom. Secondary mirror near the top. Hole in the side for the focuser. Done.

It does not work like that.

The first designs had problems I didn’t even know existed. Optics pinching, where the mirror cell holds the primary too tightly and the glass deforms under pressure, producing potato-shaped stars. Light leaks at every seam, putting gradients across my images. Wrong secondary size, cutting off the light cone and vignetting the corners of the sensor. Baffles in the wrong positions, or no baffles at all, letting stray light wash out faint nebulosity.

Each problem sent me to Google, then to books, then to forums like Cloudy Nights. I’d learn what caused the issue, go back to Fusion 360, redesign that piece, reprint it, reassemble, take it back outside, shoot, and analyze the results. Then find the next problem. It was a loop: design, print, shoot, diagnose, research, redesign.

That loop is where all the real learning happened. You can read about baffle placement in a textbook, but you don’t really understand it until you see the gradient in your image and trace it back to a light path you didn’t block.

The design process

The workflow I landed on starts with math, not CAD.

Newt for the Web is a tool that lays out the optical geometry of a Newtonian telescope. You plug in your primary mirror diameter, focal length, desired illuminated field, and it calculates everything: secondary mirror size, secondary offset, tube diameter, focuser position, baffle placement. You could do this with pencil and paper, or directly in Fusion 360, but the math is already solved in a nice-to-use tool and it works well.

Once I have the optical geometry locked, I bring those measurements into Fusion 360 and start designing the structure around them. The optical path is sacred. Every other decision, tube wall thickness, section joints, spider vane geometry, attachment points, serves the optics.

A few design principles I’ve learned:

Carbon fiber tubes for rigidity. All of my designs use carbon fiber tubes running along the circumference, in-line with the optical axis. These are small-diameter CF tubes (13x11mm on the Squat) that connect the major printed sections together. They add enormous stiffness for almost no weight. The printed sections are the structural nodes. The CF tubes are the bones.

Avoid aluminum. It’s heavy and it reacts to temperature. Aluminum expands and contracts more than you want in an instrument that needs to hold alignment across a 20-degree temperature swing over a night of imaging. Carbon fiber’s thermal expansion coefficient is near zero. PET-CF printed parts are better than aluminum too.

Print orientation matters. FDM printing is anisotropic. A part is strong along the layer lines and weak across them. Every structural component needs to be oriented so the load paths align with the layer lines, not against them. A spider vane printed flat is rigid. The same vane printed on end will snap.

Tolerances are generous. FDM printing at 0.2mm layer height gives you maybe 0.1-0.2mm dimensional accuracy. That’s fine for structural parts. It’s not fine for precision interfaces like the focuser seat or the mirror cell contact points. Those areas get designed with adjustment mechanisms, set screws and push-pull collimation bolts, to compensate for what the printer can’t provide.

The spider problem

The spider holds the secondary mirror in the light path. It’s a set of thin vanes that connect the secondary to the inside of the tube. Simple concept. Surprisingly hard to get right in a printed design.

The 114mm uses a small secondary mirror. Small enough that the spider vanes can be thin and still hold collimation. Lightweight secondary, thin vanes, minimal diffraction spikes in the final images. It just works.

The 150mm was a different story. To project a large enough image circle to cover an ASI533 sensor, the secondary mirror has to be bigger. Bigger mirror means more weight. More weight means the thin spider vanes from the 114mm design flex, and flexing means you lose collimation. The secondary shifts, your stars become elongated, your image is ruined.

I went through multiple spider redesigns on the Big Guy. Thicker vanes, different attachment geometry, testing how much the secondary shifted under different orientations as the scope tracked across the sky. Every change to the spider affects diffraction patterns in your images (those cross-shaped spikes you see on bright stars), so you’re balancing structural rigidity against optical artifacts. More vanes or thicker vanes means more stable, but more prominent diffraction spikes.

Material selection

I landed on PET-CF (carbon fiber filled PETG) for the structural components. Three reasons.

Stiffness. The carbon fiber fill makes the printed parts significantly more rigid than standard PETG or PLA. A telescope tube that flexes under gravity is useless. As the scope tracks across the sky and changes orientation, any flex in the structure shifts the optics and ruins your alignment.

Temperature stability. You’re imaging at night, often for hours. Temperatures drop. PLA’s glass transition temperature is around 60C, which sounds fine until you realize that dimensional changes start well below that point. PET-CF handles thermal cycling better. Less expansion, less contraction, less shift in your optics across a session.

Dimensional stability. The parts need to maintain their shape over months of use. PLA creeps under sustained load. PETG is better. PET-CF is better still. When your mirror cell has to hold a primary mirror at a precise angle indefinitely, material creep is the enemy.

Print settings matter as much as material choice. Layer height, infill percentage, wall count. Every part is a tradeoff between weight and rigidity. A heavier telescope is more stable but harder on your mount. A lighter one tracks better but might flex. I spent a lot of time testing different combinations, printing test pieces, loading them, measuring deflection. Not glamorous work, but it’s where the engineering actually happens.

Stray light

Stray light is photons that reach your sensor without bouncing off the primary mirror first. Light leaking through seams in the tube, reflections off the interior walls, anything that isn’t the signal you’re trying to capture. In astrophotography, you’re collecting photons from objects so faint that your eyes can’t see them. Any stray light destroys your signal-to-noise ratio.

For the 114mm, I flocked the interior. Flocking is a felt-like adhesive material that absorbs stray light. Line the inside of the tube with it, and reflections off the interior walls drop to nearly nothing. Simple, effective, no design complexity.

The 150mm needed more. I used an online baffle calculator to determine the diameter and placement of internal baffles, rings that protrude from the tube wall to block light paths that would otherwise reach the sensor. Calculated baffles plus flocking on the interior surfaces.

Even with proper baffles, 3D printed tubes have seams. Every joint between printed sections is a potential light leak. Both scopes have gaff tape covering every seam as a last line of defense. It’s not elegant, but a single light leak can put a gradient across your entire image that’s difficult to remove in processing.

Collimation

Collimation is aligning the optical elements so light from the primary mirror converges at the correct point on your sensor. In a Newtonian, that means the primary mirror’s focal point, the secondary mirror’s position and angle, and the focuser all need to be precisely aligned.

3D printed structures need regular recollimation. The tolerances in FDM printing are large by optical standards. Temperature changes, transport vibration, even gravity shifting as the scope points to different parts of the sky can knock things slightly out of alignment.

This is a known tradeoff. Commercial telescopes with machined aluminum tubes hold collimation better. But collimation takes about five minutes with a laser collimator, and you check it at the start of every session anyway. For me, it’s just part of the setup routine. You accept a few minutes of alignment work in exchange for a telescope that cost a fraction of the commercial equivalent and that you can modify anytime you want.

The imaging chain

The telescope is just one piece. The full setup has evolved a lot since the iPhone photos in Arizona.

The OpenAstroTracker with a mirrorless camera was the first real attempt. It works for wide-field shots, but tracking accuracy limits you. For deep-sky imaging through a telescope, you need better than 1 arcsecond of tracking accuracy, and I could never quite get the printed harmonic drive tuned to that level. I also built a 3D printed harmonic drive mount based on a thread on Cloudy Nights. Same problem. Wind gusts had an outsized impact on tracking. A strong gust during a 60-second exposure means that frame is garbage.

The current setup:

The harmonic drive mount was the biggest upgrade. Sub-arcsecond tracking without autoguiding, and with the guide scope running, it’s rock solid. No more lost frames to wind gusts. The difference between fighting your mount and trusting it is the difference between getting 40% of your frames and getting 95%.

That’s where I eventually accepted that some things are worth paying for. I tried to DIY the mount. Twice. The precision required for consistent sub-arcsecond tracking over hours of imaging is beyond what FDM printing can deliver. The scope is where the creative engineering lives. The mount is where you buy the precision.

Why design your own

There are existing 3D printed telescope designs you can download and print today. Good ones. If your goal is to take pictures of the night sky, printing someone else’s proven design is the fastest path. Or hell, just buy one. A commercial 6” Newtonian astrograph will do everything my designs do, with better collimation stability, for a predictable price.

Designing your own is a different thing entirely. It’s not about saving money (you won’t, once you count filament, failed prints, and the hours). It’s about whether you find the engineering as satisfying as the images.

Some people want to build the instrument. Some people want to use it. Neither is wrong. But the type of person who designs their own telescope is the type who wants to understand why the spider has four vanes instead of three. Who wants to iterate on baffle placement and see the difference in their signal-to-noise ratio. Who gets as much out of solving a collimation problem in CAD as they do out of a finished stack of the Rosette Nebula.

I’m a tinkerer. I wanted to understand every decision in the optical path, not just assemble someone else’s. And when something doesn’t work, when collimation won’t hold or a light leak shows up as a gradient in your image, you know exactly where to look because you designed every part.

The images on this site came through these telescopes. A $60 mirror, some PET-CF filament, a roll of gaff tape, and a lot of iterations. The Horsehead Nebula doesn’t care what your telescope cost.