DWARF 3 Filter Test: 7nm vs 30nm on the Eastern Veil Nebula

Why Add a Front Mounted Filter to a Smart Telescope

The DWARF 3 is a closed system by design. Everything from focus to filter selection happens inside the unit, which is a large part of why it works as well as it does for grab and go imaging. That same closed design raises an obvious question once you have used the built in filters long enough to know their limits. Is there anything beyond what ships in the filter wheel, or does the closed design mean the filter options stop where DWARFLAB drew the line.

The specific questions worth answering before spending any money: does a filter mounted externally, ahead of the stock optics, actually work on this telescope, or does the D3’s optical path reject anything not built into the wheel. If it does work, what bandpass makes sense for a smart telescope with fixed, fairly fast optics? A filter built for a slower, larger aperture setup will not necessarily behave the same way here. What size and thread standard fits a mount that has no threaded filter port from the factory. And once a filter is sitting in front of an unprotected sensor path with no enclosure around it, how is it kept from getting knocked loose, fogged, or exposed to dust during a session.

Those questions are what led to the front mounted filter build covered on this site, and this post is the direct output of that line of testing: a controlled comparison between the D3’s built in filter and a 7nm filter mounted externally, on the same target, the same night, the same exposure settings.

The DWARF 3 ships with three filters built into its internal wheel: VIS, which passes visible light with no filtering, an Astro filter for general light pollution reduction, and a Duo Band filter with a roughly 30nm passband that lets through both Ha and OIII emission together. That covers most targets well. It does not give you control over how selective the bandpass is, and being selective turns out to matter more than expected once you start comparing frames side by side.

This post covers a test built around a front mounted filter stage added ahead of the D3’s optics, using a 2 inch Optolong L-eXtreme filter with a 7nm passband, which roughly four times narrower than the built in Duo Band filter of the Dwarf 3. The filter sits in a magnetic holder built from foam board, attached using the same mounting geometry the D3 uses for its stock solar filter insert. A 3d printed version of the holder is in the works.

A Filter Replacement Along the Way

The 7nm filter used here was not the first one purchased for this project. The initial choice was a cost effective 7nm filter, and on the first session with it nothing would stack. The stacked result showed star trails across the frame, which at first reads like a tracking or polar alignment problem rather than a filter problem.

The troubleshooting step that isolated it: put the scope in Alt Az mode, mount the filter, and point at a bright star. I chose Deneb in this case. The streak appeared immediately, confirming it was not a stacking artifact.

The actual test is rotating the filter in its holder, 90 degrees is a clean rotation to use, 45 also works, and watching whether the streak rotates along with it. If the streak stays fixed regardless of filter orientation, the problem is mount or tracking. If the streak rotates with the filter, the filter itself is deviating the light, most consistent with a wedge angle in the glass or an uneven coating rather than parallel flat surfaces. The streak rotated with the filter here, which pointed to the filter as the fault rather than anything in the mount.

That filter was returned. The replacement is the Optolong L-eXtreme 2 inch 7nm shown in every frame in this post. Worth noting for anyone hitting the same symptom: a star trail on a stacked narrowband frame is not automatically a tracking issue, and the rotation test is a fast way to rule the filter in or out before spending time on guiding or polar alignment troubleshooting instead.

The Test: Caldwell 33, Same Night, Same Settings

The target is Caldwell 33, the Eastern Veil Nebula, part of the Cygnus Loop supernova remnant, cataloged as NGC 6992 and NGC 6995. Both frames were captured on the same clear, moonless night at 60 second exposures and gain 90. The Dwarf 3 built in Duo Band frame reached 1560 seconds of stacked integration, 26 minutes. The front mounted 7nm frame reached 1500 seconds, 25 minutes. The internal filter wheel was left set to Astro during the front mounted run, stacking the D3’s own broadband light pollution rejection behind the 7nm filter rather than relying on the 7nm filter alone. Both images shown here are the starless versions, so what you are looking at is the nebula’s filament structure with the stars removed from the comparison entirely.

What Changes in the Signal

A narrower bandpass rejects more of the light that falls outside the hydrogen alpha and OIII lines, which includes skyglow, light pollution, and broadband continuum light the nebula itself is not emitting. That rejection increases contrast between the nebula’s filaments and the background sky. The tradeoff is throughput. A 7nm window lets through less total light per unit time than a 30nm window, so for a fixed integration time, the narrower filter is working with a smaller signal budget. What this comparison is testing is whether the contrast gain from better rejection outweighs the throughput loss at matched integration time, on the same target, under the same sky.

What the Numbers Actually Show

Measuring both frames directly rather than relying on a visual read, using matched regions on the same filament structure in each image: the 7nm filter’s background noise floor came in roughly 40 to 45 percent lower than the built in 30nm filter, consistent with a narrower bandpass rejecting more skyglow. At the same time, the signal to background contrast in the 7nm frame was also lower, roughly half of what the 30nm filter captured in the same regions, and that loss outweighed the noise floor improvement. Measured SNR came out modestly higher for the built in 30nm filter in both of two independently tested regions, 2.39 versus 2.22 on the upper filament and 1.39 versus 0.96 on the lower hook. The two runs were not identical length, 26 minutes on the 30nm frame against 25 on the 7nm frame, a one minute difference that is far too small to account for a gap of this size on its own.

Region	Filter	Signal median	Background median	Background std	SNR
Upper filament	Duo Band 30nm	6177	5924	105.8	2.39
Upper filament	L-eXtreme 7nm	5184	5051	60.2	2.22
Lower hook	Duo Band 30nm	6098	5948	107.3	1.39
Lower hook	L-eXtreme 7nm	5128	5061	69.7	0.96

The honest read on this specific test? At roughly 25 minutes of integration, the 7nm filter is not yet paying back its throughput cost on this target. The cleaner noise floor is real and measurable, which is exactly what the physics predict, but the extra light rejected along with the skyglow appears to include enough of the nebula’s own signal that total SNR did not come out ahead at this integration length. Whether that changes at longer integration, where the 7nm’s lower noise floor has more time to compound, is a real question and not something this single test answers either way.

The OU4 Project: A Real Negative Result, and Where the 7nm Filter Fits

OU4, the Squid Nebula, has a longer history on this site than most targets. An early attempt with roughly 3 hours 20 minutes of integration showed the surrounding Flying Bat Nebula structure clearly but no trace of OU4 itself. Later attempts pushed that integration much further, eventually reaching 25 hours using the built in Duo Band filter. After a full teardown of hydrogen alpha leakage into the OIII channel, a linear FITS measurement on that 25 hour stack came back at 0.49 sigma, statistically consistent with no detection. A follow up restack at 21.9 hours, using per temperature dark calibration rather than one blanket dark library, confirmed the null result, with an estimated upper limit of roughly 1.4 percent of sky background level.

That is worth stating plainly rather than glossing over. After 25 hours of integration and careful processing, OU4 was not detected in this data, within that upper limit. This gets treated here as a legitimate negative result rather than a failed session, since a well characterized non detection is still useful information, both for this project and for anyone else attempting the same target on similar equipment.

The current run uses the 7nm filter instead of the built in Duo Band. A narrower passband is more selective around the target lines and could plausibly reduce the specific hydrogen alpha bleed that complicated the earlier attempts, independent of whether it changes total integration needed. About 7 hours are in hand so far, with two more nights planned to add another 6 to 7 hours. This project is treating 40 to 50 hours as a realistic total integration target, and this section will be updated as more data comes in rather than presenting a conclusion ahead of the evidence.

What’s Next: A 3D Printed Mount

The foam board holder works and is what produced every frame in this post, but a 3D printed version of the same mount is in development and will replace it. That will likely be its own short build post once it is finished.

Clear Skies!

---

Frequently Asked Questions