There are two very low cost smart scopes that use the same form-factor, similar optics and the same sensor. These are the ZWO Seestar S30 and the Dwarflabs Dwarf mini.
Both have 30mm objectives and focal Lengths of 150 mm at f/5. They both use the Sony IMX662 (1920x1080) sensor. The Seestar S30 presents the image in portrait format whereas the Dwarf mini presents the image in landscape format. Both scopes are primarily intended for deep sky imaging, but both are capable of imaging the Moon and the Sun (using the provided solar filters). The solar (and lunar) images produced are rather small, but because at this focal length, the image is under-sampled. During stacking, a 1.5 x drizzle can be beneficial, and produce a more useful sized stacked image.
The philosophy of solar and lunar imaging with the Seestar S30 and the Dwarf mini are completely different. The Seestar S30 captures an 8 bit RAW AVI which can be several thousand frames and the Dwarf mini captures 16 bit RAW FITS files. The Dwarf mini defaults to capturing a mere 20 frames, although this can be increased arbitrarily, but because it is a slower process, inevitably far fewer frames will be captured. The Seestar approach is more suited to stochastic or so-called 'lucky imaging' that enables the selection of the best moments of seeing during frame selection and stacking.
There are a number of ways of approaching the analysis and stacking of the FITS files produced by the Dwarf mini. However, we believe that the following is the most economical and suitable method.
We did the data capture on a very poor day through gaps in the clouds. We captured just two sets of 20 FITS files with a number of them being affected to some extent by clouds. (on a better day, larger numbers of frames could be captured to the benefit of the quality of the final image).
The gain was set to zero and the shutter speed to 1/800s which produced a correctly exposed preview on the tablet screen. (With better transparency, faster shutter speeds may be more suitable).
The 40 FITS images that were captured were placed in a single folder (directory) and Autostakkert! 4.0.13 was used to debayer and stack the best 50% of the files.
The problem is that Autostakkert! expects the origin of an image to be top left, whereas the Dwarf mini places the origin at the bottom left. The FITS Keyword "ROWORDER" in the FITS header can be set to "TOP-DOWN" (origin top left) or "BOTTOM-UP" (origin bottom left). This means that although the bayer pattern of the Sony IMX662 is RGGB, Autostakket! will by default debayer the image incorrectly. To correct this the image needs to be effectively flipped vertically, which in Autostakkert! can be achieved by forcing GBRG as the bayer pattern.
The problem is that the ROWORDER keyword is a non-standard but widely adopted FITS header extension in the amateur astronomy and astrophotography community. It is used to clarify whether an image's pixel data is written from the bottom of the image upward (BOTTOM-UP) or from the top downward (TOP-DOWN), preventing flipped images and incorrect color matrix (Bayer pattern) decoding. The rub is that neither the Seestar S30 nor the Dwarf mini write the ROWORDER into the FITS header. However, both of them write the bayer pattern into the Header. The Dwarf mini writes this: FITS Header.BAYERPAT,RGGB, FITS Header.TELESCOP,DWARF mini which is the correct bayer pattern for the Sony IMX662. The Seestar S30 on the other hand writes: FITS Header.INSTRUME,Seestar S30, FITS Header.BAYERPAT,GRBG which is not correct for the Sony IMX662. In the case of ZWO's Seestar S30, the combination of sensor orientation and their file-writing process shifts the indexing by exactly one row. A vertical shift of one row turns an RGGB pattern into a GRBG pattern. ZWO's standard driver architecture writes FITS images with a BOTTOM-UP row order. To ensure that processing programs debayer the colors correctly when loading these bottom-up files, the driver automatically translates the header's BAYERPAT string to GRBG.
The ROWORDER keyword was introduced in late 2020. While the official professional FITS standard (governed by the IAU) explicitly states that the first pixel in a FITS file should represent the lower-left corner (BOTTOM-UP), a massive influx of modern CMOS camera drivers, ASCOM, and INDI developments defaulted to writing data from the top-down. This discrepancy created massive problems for astrophotography software trying to automatically process color data.
The ROWORDER keyword was co-created and introduced by Cyril Richard and team, the developers of Siril and Patrick Chevalley of CCDCiel and Cartes du Ciel. They introduced ROWORDER (a string type keyword that takes the values TOP-DOWN or BOTTOM-UP) to allow capture software and processing software to handle image geometry seamlessly without forcing software developers to break compatibility with legacy data or calibration frames. Shortly after its introduction, other software developers—such as Han Kleijn (creator of ASTAP)—integrated and promoted its use across the amateur astronomy community. It is a shame that neither of these smart telescopes incorporate the ROWORDER keyword into their FITS headers.
The best 50% of the RAW FITS files being debayered and stacked in Autostakkert!
Cropped square in GIMP3
The stacked, cropped image being white-balanced and wavelet sharpened in waveSharp 3, software developed by Cor Berrevoets et al to replace the wavelet functions of Registax.
The final image being temperature colourised in GIMP3
The final image having been flipped vertically to its correct orientation and given a gentle sharpening in ACDSee
Final words on 1.5x drizzling in Autostakkert!
In AutoStakkert! the 1.5x Drizzle option isn't a native 1.5x pixel-drizzling math routine. Instead, Emil Kraaikamp designed it as a two-step process to maximize alignment quality while keeping file sizes reasonable. How It Works When you check the 1.5x Drizzle box in AutoStakkert! the software performs the following pipeline:
1) Native 3.0x Drizzle Stacking: AutoStakkert! takes the aligned sub-frames and executes a true 3.0x Drizzle algorithm. It shrinks the sub-pixel "drop size" and maps the data onto a grid that is 3 times the native resolution of the sensor.
2) Downsampling: Once the 3.0x stack is completely generated in memory, the software applies a Bicubic downsample to shrink the entire image by exactly 50%. The software outputs a file that is exactly 1.5x the dimensions of the original native resolution. It does this because true fractional drizzling (like 1.5x) is mathematically sloppy to compute because sub-pixel "drops" don't cleanly divide into fractional pixel grids. By performing a clean integer drizzle of 3.0x first, the software achieves much tighter, higher-quality pixel alignment during the sub-pixel stacking phase. Downsampling afterward gives the image scale required.
Even with a small amount of data captured under less than ideal conditions, it was possible to produce an acceptable image of the Sun with a Dwarf mini.
