In deep-sky astrophotography, capturing faint objects requires exposures lasting minutes. Even though modern German Equatorial Mounts (EQ-GOTO) compensate for Earth's rotation, mechanical inaccuracies in their gears (periodic error), imperfect polar alignment, and environmental factors cause microscopic deviations in star trails. The result is elongated stars and lost detail. Autoguiding is a process where a secondary camera monitors a reference star in real-time and sends software micro-corrections to the mount's motors to keep the sensor perfectly stationary relative to the sky.
The way we deliver the light of a guide star to a secondary sensor fundamentally impacts the mechanical stability and optical precision of the entire rig. Each approach addresses specific trade-offs between weight, ease of use, and susceptibility to mechanical flexure.
The setup relies entirely on the mount's built-in sidereal tracking. Any mechanical backlash or periodic error translates instantly to the image. Exposure time is severely limited by focal length and mechanical precision (typically max 30–90s).
A separate, small refractor mounted on top of the main optical tube. It offers a wide field of view, making it trivial to find a guide star. However, it suffers from a hidden enemy: differential flexure (mechanical sagging between the two tubes over time).
A tiny optical prism inserted just in front of the main camera chip redirects light from the edge of the field up into a guide camera. Because both cameras share the exact same optical tube, differential flexure is physically eliminated.
A cutting-edge integration where both the main imaging sensor and a smaller guiding sensor sit on the same PCB inside a single camera body. Offers all the benefits of an OAG without the need for complex mechanical prism adjustments or back-focus calculations.
Toggle between hardware setups to see how light travels from a celestial object, through the optics, to the camera sensors, and how each system handles mechanical errors.
Light passes through the main tube directly to the primary sensor. The mount relies solely on native sidereal motor tracking. Internal mechanical inaccuracies and worm gear flaws manifest as periodic error, stretching stars on long exposures.
When using a standard Guidescope setup, the tracking software relies on the assumption that the main telescope and the guide scope are rigidly fixed together. However, over a long exposure, gravity pulls on the heavy main camera, focusers, and the tubes themselves. This causes microscopic bending between the two systems known as Differential Flexure.
Because the guide camera doesn't see this physical bending—it only sees the stars staying centered in its own optics—the software assumes tracking is perfect. Meanwhile, the main camera's sensor is slowly sagging, creating elongated star trails. This is why Off-Axis Guiders (OAG) are heavily recommended for longer focal lengths.
Simulate a long exposure. The guide camera (green overlay) locks onto its star perfectly, but gravity causes the main optical tube to physically sag over time, destroying the main image (blue stars).
Before delving into software guiding algorithms, the physical mechanics must be addressed. An equatorial mount tracks the sky by rotating on a single axis (Right Ascension). For this to work, the mount's physical axis must be perfectly parallel to the Earth's axis of rotation (the Celestial Pole). Accurate Polar Alignment is an absolute must.
If the mount is misaligned, the target will slowly drift north or south in the camera's view (Declination drift). While autoguiding software can correct this drift by sending pulses to the DEC motor, doing so constantly forces the gears to work against mechanical backlash. Worse still, guiding on a misaligned mount introduces a phenomenon called Field Rotation. The software pins the guide star in the center, but because the mount's rotational axis is wrong, the entire star field slowly rotates around that central star, completely ruining the corners of long exposures.
Plate-Solving Method. Fast and incredibly popular. The software takes an image near the pole, prompts the mount to rotate 60 degrees in RA, takes another image, and geometrically calculates the exact center of rotation. It then provides a real-time updating target circle as you manually turn the altitude/azimuth knobs.
Three-Point Polar Alignment. A game-changer for those who cannot see the North Star (blocked by trees or buildings). It plate-solves three distinct points anywhere along the celestial equator, mathematically calculating the polar error without ever needing a line-of-sight to the actual pole.
Hardware Camera. A dedicated, wide-field camera permanently inserted into the mount's hollow polar axis. It uses proprietary software to overlay a highly magnified view of the polar region, allowing for very rapid manual alignment via a laptop screen.
The Traditional Method. The old-school, ultra-precise method. By monitoring a star near the celestial equator and another near the eastern/western horizon, it measures actual optical drift over several minutes to calculate alignment. It is slow but scientifically accurate and requires no clear view of the pole.
Use the physical Altitude (Up/Down) and Azimuth (Left/Right) knobs to align the mount's axis (Red Dot) to the True Celestial Pole (Green Crosshair). Notice how poor alignment forces the guiding software to pin the center star, causing the rest of the frame to rotate and trail over a 5-minute exposure!
The popular guiding software PHD2 faces a fundamental challenge: How do we recognize whether a star moved because of a real mechanical error in the mount (which must be corrected), or due to turbulence in the Earth's atmosphere (seeing)? If the software reacts to atmospheric turbulence, it will cause the mount to oscillate wildly (a phenomenon known as "chasing the seeing"). Therefore, PHD2 implements sophisticated mathematical filters.
| Algorithm | Primary Axis | Key Parameters | How It Works & Filtration |
|---|---|---|---|
| Hysteresis | Right Ascension (RA) | Hysteresis (%), Aggressiveness (%) | Considers the history of previous corrections. Prevents the mount from wildly overreacting to sudden spikes by averaging the new calculated correction with the historical trend. |
| Resist Switch | Declination (DEC) | Minimum Move (px), Aggressiveness (%) | Ignores minor changes and strictly refuses to issue a correction in the opposite direction unless it is certain the star has crossed the axis. Ideal for eliminating gear backlash. |
| Z-Filter | Right Ascension (RA) | Min Move (px), Exposure time factor | Uses discrete-time domain transformation to cut off high-frequency noise (rapid atmospheric shimmering). Allows only low-frequency trends—i.e., real mount drift—to pass through. |
Simulate guiding graph behavior based on the chosen algorithm, its internal parameters, and current weather conditions. Notice how incorrectly set aggressiveness during poor seeing causes the entire mount to oscillate.
Historically, guiding software locked onto a single star. If that star twinkled or warped due to an atmospheric pocket, the software erroneously assumed the mount had drifted and issued an unnecessary correction. This created high-frequency RMS noise.
Modern PHD2 introduced Multi-Star Guiding. Instead of trusting one star, the software analyzes up to 9 stars simultaneously across the sensor. Because atmospheric turbulence is highly localized and random, while one star jumps left, another might jump right. By calculating the weighted average (centroid) of all stars combined, the random atmospheric noise mathematically cancels itself out, leaving only the true, underlying mechanical drift of the mount.
Watch how atmospheric seeing causes individual stars to jump randomly. Switch to Multi-Star guiding to observe how tracking multiple stars creates a mathematically stable centroid, ignoring the noise.
Independent of software algorithms, mount precision can be radically increased by eliminating errors directly at the motor microcode level.
Classic PEC works like a recorder. The user lets the mount run through one or more complete cycles of the worm gear (usually 4 to 8 minutes). The software records repeating deviations caused by gear eccentricity and saves them to the mount's flash memory. During normal operation, the mount automatically speeds up and slows down the drive to eliminate mechanical errors before they even manifest.
A modern extension running within the PHD2 algorithm. PPEC uses mathematical models based on Gaussian processes to continuously learn the hidden sine wave of the periodic error buried beneath the atmospheric noise. Unlike normal reactive guiding, which sends a correction pulse after the star has moved, PPEC predicts gear behavior seconds in advance. It sends pulses **proactively** against the known curve of the periodic error, ensuring perfectly round stars.
Adjust the PPEC learning phase. Notice how the blue predictive curve gradually matches the red native mechanical error. When combined, the green residual error line flattens out to near-zero.
Even with perfect guiding and polar alignment, your final stacked image might be ruined by a phenomenon called Walking Noise (or Fixed Pattern Noise). Every camera sensor has microscopic hot pixels and thermal inconsistencies. If the mount tracks too perfectly, these imperfections sit on the exact same pixels across dozens of exposures.
When you align the stars in post-processing, polar alignment drift or flexure causes the image frames to shift slightly relative to the sensor. This drage the stationary sensor defects across the image, creating ugly, raining streaks.
Dithering is the software solution. Between every exposure, the guiding software intentionally nudges the mount by a few random pixels. The stars move to a new location on the sensor, but the sensor noise stays fixed. When the stacking software later aligns the stars, the sensor noise is scattered randomly and mathematically averaged out to a smooth, clean background.
Stack 30 frames to see how fixed pattern noise behaves. Notice how dithering randomly scatters the noise, allowing the stacking algorithm to completely average it out.