Motorized Zoom Lens Repeatability: 10-Cycle Test Method

This article is also available in 简体中文.

Short answer: a motorized zoom lens has no single repeatability figure — it has a magnification spread, a directional offset and a settling time, all measured on your own hardware. Lock the bench variables, drive the lens through at least ten round-trip cycles between the presets you actually use (one pass each way is one cycle), record the measured feature length at every arrival, then reduce the series to range R, standard deviation σ, maximum deviation D_max, consistency C and hysteresis H, against a threshold named before the test. N and every threshold below are engineering recommendations, not published POMEAS specifications.

Qualify the optical level, not the pulse

A motorized zoom lens is positioned by absolute pulse count and can read its position back, so qualifying it with a pulse number is tempting. Resist it: pulse is mechanical position, magnification is the optical result, and a drive returning perfectly to its commanded pulse can still image at a different scale. Measure both, but only the optical level is an acceptance criterion — with hysteresis cut out separately, as the mean arrival from above versus from below at one preset.

The published platforms used as examples

LZL-12101 is manual, so a motorized repeatability test does not apply to it — what repeats is the operator, not a drive. Its published 3.05 µm minimum resolution is the optical floor for any test run on it.

Test bench: lock these before cycle 1

Repeatability is a spread, and every uncontrolled variable lands inside it. Set each row to its production value.

Acquisition: N ≥ 10 round-trip cycles

One cycle is one full round trip, A → B → A, so ten cycles give ten arrivals at A without ever travelling the same direction twice in a row — which is what exposes hysteresis. N ≥ 10 is an engineering recommendation, not a POMEAS specification: ten is the practical floor for a meaningful standard deviation, and a tight tolerance justifies 20–30 cycles quoted with their conditions.

Read the trend, not only the total: upward drift is thermal settling, wide symmetric scatter is random play in the drive, a forward/reverse gap is backlash.

Reducing the series: five statistics

Convert pixels to magnification first: β_i = (p_i × s) / L, where p_i is the feature in pixels at arrival i, s the camera pixel pitch and L the certified object length.

Worked example (example input values, not POMEAS specifications)

Assume L = 1.000 mm, s = 3.45 µm (example value), the 1X preset of a 6.5:1 motorized platform and n = 10 alternating arrivals measuring p = 290.0 · 289.8 · 290.1 · 289.9 · 290.0 · 289.7 · 290.1 · 289.9 · 290.0 · 289.8 px:

• β̄ = (289.93 × 3.45 µm) / 1000 µm = 1.0003X
• R = 0.4 px = 1.38 µm on the sensor; σ = √(0.1610 / 9) = 0.134 px = 0.46 µm
• D_max = 0.23 px = 0.079% of β̄
• C = (1 − 0.4 / 289.93) × 100% = 99.86%
• H = |290.04 − 289.82| = 0.22 px equivalent (forward 290.0 / 290.1 / 290.0 / 290.1 / 290.0 px, reverse 289.8 / 289.9 / 289.7 / 289.9 / 289.8 px)

Here the spread is under half a pixel while the forward/reverse offset is systematic — a result that only becomes a verdict once a threshold is named.

Suggested acceptance thresholds: two tiers

No industry-wide repeatability figure exists for a motorized zoom lens. What follows is a recommended starting point for your own acceptance document — an engineering rule, not an effective POMEAS specification. Pixel-equivalent limits tie the judgement to what matters: how much of a pixel the magnification moved.

Apply the tier that matches the job, and say which one you applied. In the worked example, R = 0.4 px, σ = 0.134 px and H = 0.22 px pass the vision tier and miss the metrology tier — the lens belongs on the inspection station, not on a metrology station.

When it fails: disposition workflow

Record sheet template

| Field | Value | Field | Value |
|---|---|---|---|
| Lens model / serial | | Control line / motor generation | |
| Camera model / pixel pitch s | | Host software / SDK version | |
| Target ID / certified length L | | Certificate or serial no. | |
| Working distance set | | Focus locked (Y/N, how) | |
| Illumination / supply | | Exposure / gain (fixed) | |
| Preset ID / magnification | | Pulse value commanded | |
| Settling dwell | | Homing time | |
| Ambient temp. start / end | | Run start / end time | |
| Acceptance tier applied | Vision / Precision metrology | Run ID | |

| Run ID | Cycle | Direction (F/R) | Position read-back | Measured p (px) |
|---|---|---|---|---|
| | 1 | | | |
| | 2 | | | |
| | 3 | | | |
| | 4 | | | |
| | 5 | | | |
| | 6 | | | |
| | 7 | | | |
| | 8 | | | |
| | 9 | | | |
| | 10 | | | |

| Result | Value | Result | Value |
|---|---|---|---|
| n | | β̄ | |
| R (px / px-equiv. / %) | | σ (px / px-equiv.) | |
| Dmax (px / %) | | C (%) | |
| H (px-equiv.) | | Verdict (pass / fail tier) | |
| Operator / date | | Reviewed by / date | |

Keep the worst case seen, not only the pass result: R and H drift with wear, so store the measured scale per preset with the pulse value, and re-run the test after a camera or mount change, a focus or zoom lock adjustment, a repair, or a move to a different ambient band.

Rather than build the threshold from scratch, send us the preset positions, the tolerance and the camera pixel pitch — start at the contact page. For the mechanism behind a preset, see the pulse-to-magnification mapping and the RS-232 command set; when a preset misses, the parfocality and repeatability FAQ and the troubleshooting guide cover the fault side. The example lenses are the LZ-650104 6.5:1 motorized zoom lens and the LZL-12101 12.5X coaxial continuous zoom lens; while choosing a platform, read the magnification and FOV calculator and the note on image shift after a magnification change.

Frequently Asked Questions

How many round-trip cycles should I run to qualify a motorized zoom lens?

N ≥ 10 is an engineering recommendation, not a POMEAS specification: one cycle is a full round trip, so ten cycles give ten arrivals at the preset with the approach direction alternating every time. Ten is the practical floor for a meaningful standard deviation. If the tolerance you inspect against is tight, run 20 to 30 cycles, and always quote N next to the result — a figure without its cycle count and conditions cannot be compared with anything.

What exactly is "magnification consistency", and how is it calculated?

It is the fraction of magnification preserved across the test series: C = (1 − R / β̄) × 100%, where R is the range between maximum and minimum measured magnification and β̄ is the mean. It is a convenient single number for a specification column, but not a substitute for the raw series — the same C can hide a large σ with no drift, or a smaller spread with a strong directional offset. Publish C together with N, σ, H and the presets tested.

Do I need a laser interferometer or a special target to run this test?

No. A certified target with a feature long enough to fill most of the field plus the production camera is sufficient — the 0.01 mm grid target used for centering work is a reasonable starting point. Record the certified length and certificate number, freeze exposure and gain, disable auto-exposure and auto-gain, and compute the pixel-to-millimetre scale at the preset under test. The scale belongs to the position, not to the lens, so never inherit it from another preset.

Is the spread I measure a defect in the lens?

Usually not. Most magnification spread comes from conditions: an uncontrolled ambient temperature, a brightness change that moves the edge threshold, too short a settling dwell, or a focus that was never locked. Compare the position read-back spread with the optical spread before concluding anything. Tight read-back with a loose image points to the optical or mounting side; both loose points to the drive or the approach direction. Only after those are cleared is it worth discussing the lens itself.

What is the difference between repeatability and hysteresis, and does hysteresis matter?

Repeatability is the scatter of arrivals at one preset; hysteresis H is the offset between the mean arrival from above and the mean from below. They need different fixes. Hysteresis is normally solved by process rather than hardware: adopt a unidirectional approach rule so every preset is reached from the same side. That single change often removes most of the spread with no mechanical work, which is why the acquisition procedure alternates direction — otherwise the two effects are added together and neither can be read.

Can I compare my result with a published specification?

Not directly. POMEAS publishes optical figures for these platforms — the coaxial 12.5X platform is specified at 3.05 µm minimum resolution and 0.001% minimum TV distortion — but a repeatability figure is conditional on temperature, illumination, settling time, cable length, travel speed and approach direction, so it must be measured on your own hardware. Those published figures do set the optical floor: asking the test to resolve below what the optics can render means you are measuring noise. Send us your preset positions and pixel pitch and we will help you set the tier.