Vibration & Frame Blur

Why Vibration Is Different

Most focus stacking artifacts are systematic. Haze follows predictably from the optics of defocused foreground structures, and magnification breathing follows from the geometry of a camera that travels toward its subject. Both can be anticipated, measured and planned around. Vibration is different: it is random, it varies frame to frame, and it corrupts the raw material the stacking algorithm works from rather than the composite it produces.

A single blurred frame in a deep stack does not simply contribute one missing depth slice. The stacker selects the sharpest pixel at every position across every frame, so a vibration-blurred frame can inject softened edges and phantom detail into regions of the composite where it happens to win that comparison. The damage lands where you did not expect it, and it lands in the one place you cannot reach afterwards.

Worst of all, vibration is effectively invisible at the moment of capture. Individual frames often look acceptably sharp at normal viewing magnification. The composite is where the accumulated effect of sub-pixel displacement across dozens or hundreds of exposures finally becomes visible — by which time the specimen has moved on.

The Scale of the Problem

At 5× magnification on a Micro Four Thirds sensor, one pixel samples roughly 0.75 µm of the subject plane. A physical displacement of just one or two micrometres — a shutter curtain settling, a rail stepper pulse, a footstep in the next room — translates to one to three pixels of blur in the captured frame. That is enough to render trichome tips, sepal margins and pollen ornamentation unresolvable, and at 10× the same displacement costs proportionally more.

This is not conventional camera shake. A blurred handheld frame is discarded and reshot; in a focus stack, every frame is the only record of its depth slice. You cannot redo frame 147 of 229 — the flower has continued living, breathing and wilting since it was taken.

Where Vibration Comes From

For a mirrorless stacking setup, four sources dominate, in rough order of how much damage they do:

1. Mechanical shutter actuation. On DSLRs the mirror flip is the largest single vibration event in the whole capture sequence, and its shock is transmitted directly through the body into the lens and rail. Mirrorless cameras eliminate it — but the mechanical focal-plane curtain still produces a measurable impulse at the mid-range shutter speeds (roughly 1/8 s to 1/125 s) that continuous-light macro work tends to use.
2. The rail's stepper motor. Every step delivers a short mechanical pulse into the platform, and without a settle delay between rail motion and exposure that pulse contaminates the very frame you just advanced to take.
3. Floor-borne transmission. HVAC plant, foot traffic, road traffic and nearby machinery travel through the building structure into the table legs and up into the camera. These vibrations are low frequency — roughly 2–20 Hz — and continuous, which makes them uniquely awkward: they are present during the exposure itself, so no settle time can wait them out.
4. Air movement. Draughts from vents, convection from lighting, and opening and closing enclosure panels sway stems and pedicels. On a flower whose pedicel is only a few millimetres long, sub-millimetre movement at the tip is a multi-pixel displacement at 5×.

A Control Hierarchy That Works

Because the sources stack on top of each other, control has to as well. In order of return on effort, for a modern mirrorless body:

1. Electronic shutter, always. It is the single most effective mitigation available in the camera body — it removes mechanical shutter vibration entirely. On focus-bracketing drive modes it engages automatically; for manual single-frame capture during setup it usually has to be selected explicitly.
2. A settle delay of at least two seconds between rail motion completing and the shutter firing. Two seconds is long enough for the stepper pulse to decay at typical mounted masses, and short enough not to dominate the session on a 200-frame stack.
3. Mass and isolation under the whole assembly. A heavy, layered tabletop on vibration-isolation feet attenuates floor-borne transmission in exactly the frequency range where it lives. The heavier the better, within reason — mass is cheaper than a failed deep stack.
4. An enclosure that stays closed during capture. It suppresses air movement around the specimen, slows desiccation of living material, and blocks the draught you cause every time you check the framing.
5. Thermal and mechanical settling. Allow the system half an hour to reach thermal equilibrium before the first stack, and a full five minutes of stillness after any disturbance — repositioning the rail, changing the specimen, opening the enclosure.

Recognizing Vibration Damage in a Composite

Vibration contamination in a finished stack shows up in three recognizable ways. Ghost edges: a fine structure — a trichome, a vein edge, a sepal margin — appears doubled or softly echoed in the composite, yet is sharp in at least some of the individual frames. Sharpness bands: strips of reduced detail running across the image at specific depth ranges, marking where the affected frames contributed. And generalized softness that is not localized around foreground structures — which is the tell that separates it from haze.

Diagnosis is manual but reliable: step through the source frames in sequence and compare each against its neighbours. Frames that look softer than the ones immediately before and after identify the exposure events that were contaminated, and their positions in the sequence tell you which depth ranges of the composite to distrust.

What Can Be Recovered

If the affected frames are few and the overlap between slices is generous, dropping them from the stack and re-processing may cost nothing. Where coverage does not allow it, the affected region can be retouched from a neighbouring depth slice — workable, but slow, and always a compromise between sharpness and structural accuracy.

Neither option restores what was lost. Prevention is the only strategy that actually pays: the control hierarchy above takes minutes to implement, costs a fraction of one ruined deep stack, and removes the one artifact class that cannot be repaired after the specimen is gone.

Discuss a stacking workflow →