Star-Stacking the Milky Way With a Canon EOS R5 Mark II

I spent three clear nights last October on a ridge above the Alvord Desert with a Canon EOS R5 Mark II I'd owned for about six weeks, mostly to answer one question for myself: does the newer 45-megapixel stacked sensor actually change anything about star stacking, or is it just a resolution bump wrapped around the same body I already knew? The honest answer is that it changes the noise math more than the megapixel count does, and that matters a lot more than a spec sheet number once you're stacking dozens or hundreds of frames to pull a clean sky out of the dark.

This isn't a camera review. It's what I actually do with this body on a stacking night, from setting exposure times that respect the pixel pitch, to picking a tracker that won't tip in wind, to the point where I stop shooting subs and start sorting them, which for me now happens partly inside imagic before anything touches a stacking program.

What the stacked sensor actually changes at night

The R5 Mark II uses a 45-megapixel stacked, back-illuminated sensor, and the part that matters for night work isn't the resolution figure, it's the dual conversion gain circuit. On the original R5 I shot before this one, noise cleaned up meaningfully once you crossed into the ISO 800 to 1600 range, where the sensor switches to a lower-noise gain mode. The Mark II carries the same behavior, and in practice I now shoot most wide-field nightscapes around ISO 3200 rather than pushing to 6400 or 12800, because the extra stop or two of clean shadow detail from stacking twenty subs outweighs anything I'd gain from one hotter exposure.

Smaller pixels do mean more read noise per pixel than a 24-megapixel body shows you at 100 percent, but that's exactly the problem stacking solves. Average thirty subs together and the random noise per pixel falls by roughly the square root of the frame count, while anything that lines up frame to frame (the stars, if you're tracking, or the ground, if you're not) stays sharp. What you do feel is the file size everywhere else in the pipeline. A single night can produce 60 to 150 gigabytes of CR3 raws once you count bracketed foreground frames and focus tests, and that volume, not the sensor, is the actual bottleneck on a laptop parked at a trailhead at 2 a.m.

Glass that keeps coma under control

Wide apertures at the edge of the frame are where cheap lenses fall apart on a star field, since coma turns points of light into little wings or seagulls in the corners. The RF 15-35mm f/2.8L is what lives on my camera most nights. It's not perfectly clean wide open in the far corners, but stopped to f/4 the stars stay round almost to the edge, and 15mm gives enough headroom to shoot a longer single exposure than 24mm would allow before trailing shows up. The RF 14-24mm f/2.8L is the sharper option at the extreme wide end and genuinely excellent for corner-to-corner star points, but it's heavy, expensive, and I only reach for it on the nights I know the Milky Way arch is the whole point of the trip.

For a budget setup, an old Rokinon (Samyang) 14mm f/2.8 adapted from EF still earns its keep. It has more visible coma than either RF lens wide open, but stopped down to f/4 it's usable, and it costs a fraction of the native glass. Autofocus doesn't matter here anyway; every one of these lenses gets focused manually with focus magnification on a bright star, not left on AF.

Exposure time before you even think about tracking

The old "500 rule" (500 divided by focal length equals your max shutter speed in seconds) was written for lower-resolution sensors, and it falls apart fast on a 45-megapixel file. At that pixel density, star trailing that would be invisible on a 20-megapixel body shows up clearly the moment you zoom to 100 percent, which is exactly what happens the first time someone stacks a set of R5 Mark II subs and finds the points have gone slightly oval. In the field I use something closer to a 300 rule as a starting ceiling, then check a test frame at full resolution before committing to a sequence.

These aren't laws of physics, they're field numbers from checking test frames at 100 percent on this specific body. If you're stacking rather than shooting a single hero frame, you can actually afford to be a little more conservative than a one-shot photographer would be, since you're going to average the noise down across many subs anyway. There's no reason to gamble on trailing to save a stop of ISO.

Running the interval timer

The built-in interval timer under the shooting menu handles a full stacking sequence without an external remote. For an untracked nightscape stack I set 40 to 60 frames at whatever shutter speed the table above allows; for a tracked deep-sky target I've let sequences run past 200 frames. Two settings matter more than people expect: turn off long exposure noise reduction before you start, because it doubles the time between subs and can eat your whole shooting window without adding anything a calibration frame won't fix better, and if it's a cold night, plug in USB-C power delivery so the camera charges while it shoots. The LP-E6P drains faster than you'd guess once the temperature drops below freezing and the screen stays on between frames.

Tracked versus untracked: where a star tracker earns its keep

For wide nightscapes with a real foreground in the frame, I don't track. I shoot the sky stack at the exposure limits above, shoot the foreground separately at blue hour or with a light, and blend the two later. A tracker doesn't help here, since it would blur the ground the moment it starts moving.

For anything longer than about 35mm, or for actual deep-sky targets like a nebula or a galaxy core, a tracker changes what's possible. Mounted on a Star Adventurer GTi or an iOptron SkyGuider Pro, the same body can run 60 to 120 second subs at 85mm or 135mm without star trails, because the mount is following the sky's rotation instead of fighting it. The tradeoff is payload. The R5 Mark II plus an RF 100-400mm or even a compact RF 135mm pushes close to what these smaller trackers are rated for, and an unbalanced payload shows up as periodic tracking error in your subs, little wobbles that no amount of stacking fixes. I counterweight deliberately now rather than eyeballing it, and I check balance in both axes before I walk away from the tripod for an hour.

Picking a stacking program

The camera and the tracker only get you the subs. What actually turns forty or two hundred raw frames into one clean image is the stacking software, and the right pick depends on whether you're doing a tracked deep-sky stack or an untracked nightscape.

Where imagic fits into a night like this

A single night of star stacking produces hundreds of near-identical frames, plus a scattering of focus tests, foreground brackets, and the occasional wind-shaken sub that looked fine on a 3-inch screen and isn't. Scrolling through all of that by eye at 1 a.m. is miserable, so I run the folder through imagic's local sharpness scoring before I open a stacking program at all, and its duplicate and burst clustering groups the long runs of nearly identical star subs so I'm not scrolling past forty versions of the same frame one at a time. If you want the mechanics behind how that scoring actually works, I wrote about it separately in how AI photo culling works.

One honest caveat: sharpness scoring built around daylight subjects and faces behaves a little differently on faint point sources against a black background, so I still spot-check a few frames from each cluster before I trust the group wholesale, especially the ones near the start and end of a sequence where wind or dew are most likely to have caught a frame mid-gust. And because it's all local processing with nothing uploaded anywhere, it runs fine on a laptop with zero signal parked at a trailhead, which matters more than it sounds like it should when you're two hours from the nearest cell tower.

Blending the stack back into one photo

Once the sky is stacked in DSS or Sequator, the real work is matching it to a separately exposed foreground. Blue-hour foreground shots and a stacked night sky almost never share the same white balance out of the camera, and pulling them into agreement by eye, frame after frame, across a season of shoots, is where a lot of astro edits start to look inconsistent from one image to the next. I keep a preset baked from my own edited examples through imagic's apply_my_style feature, so the color treatment on a new stack starts from something that already matches how I've been grading this kind of shot rather than from a blank slate. It doesn't replace the blend itself, which still happens by hand with layer masks, but it saves the fifteen minutes I used to spend re-deriving a white balance curve I'd already landed on three shoots ago. For the actual mechanics of matching a warm foreground to a cool sky, this color grading guide covers the split-toning approach I use.

Mistakes I made so you don't have to

• Leaving long exposure noise reduction on during a 200-frame interval sequence, which doubled the total shoot time and cost me the last hour of a clear window before clouds rolled in.
• Trusting the rear LCD preview at 100 percent zoom on a dimmed screen, then finding real star trailing on the laptop the next morning. Now I bump screen brightness up for every focus check, even though it kills the battery faster.
• Skipping flat frames on a tracked sequence and getting a stubborn vignette that got worse, not better, once the stretch in PixInsight amplified it.
• No dew heater on a humid coastal night. The front element fogged over about ninety minutes in, and the back half of a 150-frame sequence was unusable before I noticed on a rear-screen check.

Frequently Asked Questions

Do I actually need a star tracker for the R5 Mark II, or is the sensor good enough without one?

For wide nightscapes with the Milky Way and a foreground in frame, no, you don't. The exposure limits in the table above plus a stack of thirty or so subs get you a clean result without a tracker. A tracker only becomes necessary once you want longer focal lengths or actual deep-sky targets, where minutes of exposure per sub are the point.

What ISO should I shoot at for a star stack on this body?

ISO 3200 is where I land most nights, since it sits above the sensor's dual gain switch without pushing so hot that highlight clipping on bright stars becomes a problem. I'll go to 6400 on genuinely dark, moonless nights with a fast lens if I'm shooting fewer subs than usual and need more signal per frame.

Will 45-megapixel raw files slow down my stacking software?

Noticeably, yes, especially DeepSkyStacker and PixInsight on a sequence past a hundred frames. Stacking two hundred R5 Mark II raws can take well over an hour on a mid-range laptop. I now do a first sharpness pass and remove the obvious misses before the files ever touch the stacking program, which cuts real time off that step.

Can I mix subs shot on different nights into one stack?

You can, but only if the framing, focal length, and focus point genuinely match, which is harder than it sounds without a tracker holding a fixed target. For untracked nightscapes I don't bother, since even a slightly different tripod position shifts the horizon enough to fight the software's alignment. For a tracked deep-sky target locked to the same coordinates, combining sessions across a few nights with matching calibration frames works fine and is honestly the only practical way to get faint nebulosity out of a light-polluted backyard.