Plenty of instruments image the same scene through more than one filter. A fluorescence reader cycles through excitation or emission bands. An inspection station checks a part in several color bands. An observatory measures a target's color, or a Space Domain Awareness (SDA) site builds light curves in two bands. In every case, the time spent changing filters is time the sensor isn't collecting light.
With short exposures, the filter change can take longer than the exposures themselves. This post covers how to budget that time, how to keep the wheel and camera in step, and what to settle before you choose the hardware. We use Finger Lakes Instrumentation (FLI) cameras and our new High Speed Filter Wheel as examples.
Where the time goes in a multi-band cycle
One full multi-band cycle has four parts for each filter:
- Move the wheel to the filter.
- Let it settle.
- Expose.
- Read out the frame and hand it to the host.
Some of these can overlap. Readout of one frame can often run while the wheel moves to the next filter, as long as the sensor isn't integrating during the move. Others can't overlap, and the slowest one sets the pace. When exposures are long, the move time barely registers. When exposures are a few milliseconds, as in bright-field inspection, fast fluorescence work or daylight sky imaging, the move can dominate the cycle.
Budget the move time
Start with a simple sum. Take the number of filters in the cycle, multiply by the move time per step, and compare that with the total exposure time per cycle.
Here's an example with round numbers. Say you cycle through four adjacent filters with 10 ms exposures, so 40 ms of light per cycle. Ignore the return move for now; we come back to it below. A wheel that takes about 1 s per filter change (typical of a general-purpose stepper wheel, including FLI's CenterLine wheels) adds about 4 s of moves, so the sensor collects light for about 1% of the cycle. A wheel that moves to the adjacent filter in 30 ms adds about 120 ms, and the share of the cycle spent collecting light rises to about 25%.
The FLI High Speed Filter Wheel moves to the adjacent filter in 23 ms on the HS-625 and in 30 ms on the other models, and its longest move takes 90 ms. Run the same sum with your own exposures and band count before you decide whether wheel speed matters for your instrument. For long, faint-target exposures, a standard wheel is often fine. Our accessories page lists FLI's standard wheels.
Never expose while the wheel moves
A frame captured while the filter is still moving mixes two bands and can show a moving edge across the field. The fix is strict sequencing: the camera starts integrating only after the wheel reports or is known to have finished its move and settled.
How you enforce that depends on the camera. FLI Aurora cameras have external trigger in and out as standard. Kepler cameras support hardware triggering through the SDK. Aurora isn't offered with a mechanical shutter, so its exposure is set electronically, while Kepler cameras support internal shutters. Either way, plan the timing so integration starts after the move. Measure the real delay between the move command and a stable image on your bench rather than relying on the nominal move time alone.
Let the hardware run the sequence
Software-timed sequences add latency and jitter. The host sends a move command, waits for the wheel, triggers the camera, waits for the frame, and repeats. Each round trip through the operating system adds time, and that time varies.
A stored sequence takes the host out of the loop. The High Speed Filter Wheel can store a sequence of up to 50 filter moves and step through it with a trigger input, so the wheel keeps pace with your exposures without waiting on the host computer. A common arrangement uses the camera's trigger output, or a single timing source driving both devices, to advance the wheel after each frame. If several cameras share the cycle, drive them from an external sync or trigger signal too, since independently timed cameras drift.
FLI's CenterLine wheels need a host computer and have no stand-alone or trigger-input control, so they suit slower, software-driven cycles.
Order your filters for short moves
Move time depends on distance. On the High Speed Filter Wheel, an adjacent step takes 23 to 30 ms, while the longest move takes 90 ms. Load the filters in the order you'll use them, so most of the cycle is adjacent steps.
A cycle that runs 1, 2, 3, 4 and then jumps back to 1 pays the long move once per cycle. Some instruments avoid that by running forward and then backward (1, 2, 3, 4, 3, 2, 1). Others simply accept one longer move as part of the budget. Either choice is fine once it's counted.
Keep focus across filters
Filters of different thickness shift the focus by different amounts. A flat glass plate in a converging beam pushes focus back by about a third of its thickness, for typical glass. A wheel that mixes a 2 mm filter and a 5 mm filter shifts focus by about 1 mm between them, which matters in a fast optical system.
The simplest fix is filters of matched thickness, or parfocal sets. Otherwise, adjust focus per band, or design the depth of focus to cover the difference. The High Speed Filter Wheel accepts filters 1 to 9 mm thick. Our post on back focus, flanges and filter wheels covers spacing in detail.
Matching the camera to the cycle
A fast wheel helps only if the rest of the chain keeps up. Check three things:
- Readout and transfer. Make sure the camera, interface and host can move each frame out before the next exposure ends. Plan for continuous transfer to the host.
- Trigger timing. Measure the camera's response to a trigger in the mode you'll actually use.
- Signal per band. Short exposures in narrow bands may collect little light. Confirm signal-to-noise in every band, not just the brightest. Our post on acceptance testing for gain, noise and linearity shows how to measure what you have.
For a life-science instrument, see our post on choosing an OEM camera partner. For photometry and SDA timing, see our posts on time-series photometry and daytime satellite tracking.
Questions to settle before you choose
- How many bands are in each cycle, and what exposure does each need?
- What share of the cycle can you afford to lose to filter moves?
- Will the host time the sequence, or will hardware triggers?
- Which device leads: the camera, the wheel or an external timing source?
- What filter diameter and thickness will you use, and are they parfocal?
- How fast must each frame reach the host, and how much data does a full run produce?
Talk to our sales team
If you're building a multi-band instrument or speeding up an existing one, send our sales team your exposure times, band count, filter size and thickness, and the camera you're using or considering. We'll help you match a High Speed Filter Wheel or standard wheel to an Aurora or Kepler camera. FLI cameras and the High Speed Filter Wheel are Made in USA. Request a quote.
