A focus stack of a sub-millimeter snow crystal usually fails for one of two reasons. The crystal sublimates while you're shooting, or the metal rail shrinks in the cold and shifts your focus plane by more than one step. Most stacking tutorials mention neither.
You've probably had a 90-frame stack come out of Helicon looking like a watercolor. The dendrite tips have halos, the centre hexagon is sharp, and the edges look smeared. Then you reshoot with more frames and smaller steps, and it gets worse.
What follows are the numbers I use: step sizes tied to the actual depth of field at each magnification, time limits for each stack, and the thermal routine that stopped my rails from drifting halfway through a sequence.
Why Most Focus Stacking Advice Fails on Snow Crystals
The usual advice is to take 10–20 frames, either by turning the focus ring or with an automated rail, then merge them in Photoshop or Helicon Focus. That works for a beetle on a leaf at 1x in a warm room. On a 0.6 mm stellar plate, it causes three problems:
- Turning the focus ring changes magnification. Each frame comes out at a slightly different scale, and the alignment step has to correct for this. Fine dendrite branches don't survive that correction well, so they come out soft or doubled.
- Ten to twenty frames is a number someone guessed. At 3x the usable depth of field is in the tens of microns. A crystal tilted even slightly on its fibre needs far more frames than 20 to cover it.
- Nobody mentions that the subject is shrinking. A snow crystal changes from solid to vapor at temperatures well below freezing. Its tips are the thinnest ice on it, so they disappear first.
My first serious season went into finding that third problem the hard way. Stacks looked fine at 1x and fell apart at 3x. I assumed the problem was vibration and spent a month adding damping. It wasn't vibration. The tips were literally getting shorter between the first frame and the last.
How to Read the Step and Speed Parameter Matrix
I set up every session from this table. The depth of field figures come from the standard macro formula: DOF ≈ 2 · N · c · (m+1) / m². I used a circle of confusion of about 9 µm, which fits a high-resolution full-frame sensor.
| Magnification Ratio | Lens Aperture (f-stop) | Mathematical Depth of Field (µm) | Max Step Size Setting (µm) | Recommended Stack Capture Time Limit |
|---|---|---|---|---|
| 1x (life size) | f/8 | 280 | 200 | Under 180 seconds |
| 2x | f/5.6 | 85 | 60 | Under 120 seconds |
| 3x | f/4 | 32 | 20 | Under 60 seconds |
| 5x | f/2.8 | 11 | 8 | Under 30 seconds |
Each maximum step size is about 65–75% of the depth of field, so every slice overlaps the next. Treat that column as the absolute limit, not the target. At 3x I actually shoot at 12 µm steps, for reasons I'll get to in the rail drift section.
Why Every Row Ends Up Near f/16 Effective
Calculate the effective aperture for each row with N × (1 + m):
- 1x at f/8 → f/16 effective
- 2x at f/5.6 → f/16.8 effective
- 3x at f/4 → f/16 effective
- 5x at f/2.8 → f/16.8 effective
That's deliberate. Around f/16 effective is where diffraction softening starts to beat any depth you gain on a modern sensor. A lot of people stop down to f/11 at 3x because that feels like normal macro practice. That's f/44 effective, and the whole image turns to mush. If your 3x frames look soft even at the plane of focus, check this before you blame the lens.
How to Work Out Frame Count Before You Start
Measure the crystal's depth along the lens axis, not its width. A flat plate lying perfectly square to the lens might only be 50–100 µm deep. The same plate tilted 15° on a wool fibre can be more than 200 µm deep.
My formula: frames = (subject depth ÷ step size) + 6. The extra six are buffer frames, three before and three after.
Check the result against your time limit. At 3x with 12 µm steps and 2 frames per second, you can shoot 120 frames in 60 seconds, which covers about 1.4 mm. That's plenty. At 5x with 8 µm steps, 60 frames in 30 seconds only covers 480 µm. That's why I don't bother taking tilted crystals to 5x. I level them first or drop to 3x.
How to Shoot Faster Than the Crystal Sublimates
In my sessions at around −5°C, the delicate dendrite tips start visibly losing their shape after about 90–120 seconds under continuous camera light. Simple prisms and thick plates last much longer. Fernlike stellar dendrites, the crystals everyone wants, go first.
You can see the effect in the merged image. The stacking software takes the sharpest pixels from each frame, but frame 5 and frame 80 caught the tip at two different lengths. The result is a ghostly double edge, or a tip with a sharp core and a faint outline around it. People call this "halo" and try to fix it with radius settings. Radius settings can't fix a subject that physically changed during the stack.
Reduce the Heat Your Light Puts on the Crystal
Continuous LED panels are the biggest problem. Even a "cool" LED sends a steady stream of radiant energy onto a speck of ice weighing a fraction of a milligram. Here's what I changed:
- I moved to flash at 1/64 or 1/128 power. The total energy reaching the crystal over a full stack is a tiny fraction of what a continuous light delivers, and each short pulse also freezes any small movement.
- I diffuse through a cold diffuser. A frosted plastic cup that's been outside for an hour won't radiate stored warmth back at the crystal.
- I turn on a dim focusing light only while framing, then switch it off before the stack starts.
- I stopped breathing toward the set. I know that sounds obvious. A scarf over my nose and mouth was still the cheapest fix of the whole season.
Why 2 Frames per Second Is the Minimum
Flash recycle time sets your frame rate. At 1/64 power, most speedlights recycle well under half a second, so 2 fps is realistic. If your stack would take longer than the time limit in the table at that rate, there are only three ways out: bigger steps (up to the limit), lower magnification, or a crystal that sits flatter. Don't slow down to "be careful." Slowing down is exactly what lets the tips melt away mid-stack.
How to Stop Thermal Rail Drift Ruining a Stack
This is the part almost nobody writes about. Most focusing rails are aluminium, and aluminium expands and contracts by about 23 microns per metre for each degree Celsius.
Do the maths for a typical 100 mm rail travel. Say you take it from a gear bag at 15°C into −5°C air. That's a 20°C drop:
0.1 m × 23 µm/m/°C × 20°C ≈ 46 µm of contraction.
At 3x, 46 µm is more than your entire depth of field. At 5x it's almost six of your 8 µm steps. And the rail doesn't contract all at once. It shrinks steadily as it cools, which means your "fixed" step sequence slowly walks off position while you shoot. The lens barrel, extension tubes and bellows standards all contract too, each at its own rate.
Let the Rail Sit in the Cold for 30 Minutes
My rule: the rail, camera body and lens all go outside and sit for 30 minutes before the first real stack. I spend that time collecting crystals, cleaning the slide and setting the lights. In my experience, stacks shot in the first ten minutes have a noticeably higher reject rate. After the 30-minute soak, the drift mostly goes away.
A few extra habits help:
- Don't touch the rail body with bare hands between stacks. Your fingers warm the metal locally, and you'd be adding the drift back yourself.
- Leave the camera on standby rather than switching it off and on. Sensor heat stays fairly constant instead of rising and falling.
- Keep the 12 µm step at 3x even after soaking. The extra overlap absorbs any small drift that's left.
Spot Drift in the Stack Itself
Scroll through the source frames in order. If the focus plane moves through the crystal smoothly at first, then seems to stall or jump, the rail is still changing temperature. If the crystal's outline gets smaller as you go, it's sublimation. These two problems need different fixes, so learn to tell them apart.
How to Set Up Alignment in Helicon Focus and Zerene Stacker
Move the whole camera on the rail and leave the focus ring alone. Magnification then stays nearly constant, and the software has much less scale correction to do.
| Software | Method I Use | Why |
|---|---|---|
| Zerene Stacker | PMax first, then DMap for comparison | PMax holds onto fine dendrite detail but adds noise. DMap gives cleaner flat plates. |
| Helicon Focus | Method B (depth map), low radius, smoothing 2–4 | Method C (pyramid) increases contrast too much at ice edges and adds false halos. |
| Photoshop Auto-Blend | I don't use it for crystals | Masking at this scale is too coarse and loses detail along the branch edges. |
Whatever you use, plan to retouch. Zerene's retouching tool lets you paint in sharp regions from individual source frames. When a tip sublimated partway through, I paint that tip in from the earliest frame where it was sharp. That's how to save a stack that's 90% good.
Which Crystals to Stack at Each Magnification
- Stellar dendrites, 2–4 mm across: shoot at 1x–2x. They're big enough, and their tips are the fastest to sublimate.
- Sectored plates, 0.5–1.5 mm: 2x–3x is the best range. They're flat, sturdy and forgiving.
- Tiny hexagonal plates under 0.5 mm: 3x–5x, but only once the crystal is levelled.
- Columns and capped columns: these are deep along the lens axis, so 2x is the practical limit unless you're willing to shoot very long stacks.
FAQ
Can I use the focus ring instead of a rail at 2x?
You can, but expect softer branch edges. Turning the ring changes magnification from frame to frame, and the alignment correction smears fine detail. A manual micrometer rail is a better budget option than the focus ring.
Does a colder day mean slower sublimation?
Generally yes, though dry air works against you. In my experience, a crystal at −12°C in dry, still air doesn't always last longer than one at −5°C in humid air. Watch the tips, not the thermometer.
Is electronic shutter better for crystal stacks?
Yes, as long as you use flash sync limits correctly. It removes shutter shock, but many bodies limit flash use or slow down with the fully electronic shutter. Electronic first curtain is usually the reliable middle ground.
Will a carbon-fibre or steel rail drift less?
Steel expands at roughly half aluminium's rate, so it drifts less, but it still drifts. The 30-minute soak still applies. You'll just see smaller errors if you skip it.
Why do my stacks look fine on screen and soft when printed?
Usually your steps exceeded the depth of field and left out-of-focus bands between slices. These show up at 100% zoom and in large prints. Recalculate using the table rather than going by eye.
Is cleaning stacking artifacts by hand cheating?
No. Retouching from your own source frames only restores detail you actually captured. It doesn't add anything that wasn't there.
What to Do Before Your Next Snowfall
Before snow is forecast, measure how far your rail actually moves per step. Photograph a millimetre scale at 3x and check the movement against the setting. Cheap rails sometimes move less or more than the step size they display.
Then make a field card for your own setup: magnification, aperture, step size, frame count and time limit, copied from the table above. Tape it to the rail.
On the day, set up your gear outside first, start a 30-minute timer, and collect crystals while you wait. Take your first stack at 2x on a sectored plate, and save 3x and 5x for when the setup is working. Afterwards, scroll through the source frames in order. They'll show you whether you're fighting sublimation or drift.
0 Comments