Choose the homogenizer when the target is flat, the working distance is fixed and one source can supply the flux; choose a multi-lamp array when the area is large enough that no single 300 W class lamp can fill it. Both routes reach the same specification by different physics, and both must be verified at the sample plane rather than at the exit port.
Both routes deliver flux that is even to within a stated band over a stated area, but they get there differently: one remaps a single beam, the other overlaps several.
A homogenizer — an internally reflecting rod, a light pipe or a lenslet array — mixes the beam so that every point at the output receives a contribution from the whole input. The filament image, the arc hot spot and the reflector structure are all smoothed out. The cost is étendue: area multiplied by solid angle is conserved, so a larger output area forces a larger entrance aperture or a shorter working distance. The PLS-PGL-O1000 is built on this principle: one source, one homogenizer, one output plane.
A multi-lamp array takes the opposite approach: each head has its own reflector and cone, arranged so their overlap averages out the structure of any single one. Scaling means adding heads, not enlarging optics.
The differences that matter are not the uniformity numbers on the datasheet but where the light is lost, where the heat lands and how the thing fails.
| Criterion | Single source + homogenizer | Multi-lamp array |
|---|---|---|
| Source of uniformity | Remapping inside the rod or lenslet array | Overlap of several cones at the plane |
| Scaling to a larger area | Larger aperture or shorter working distance; flux density falls | Add heads at the edge |
| Edge behavior | Falls sharply at the rod boundary | Edge heads can be trimmed to match the center |
| Angular content | One well-defined cone, easy to filter or collimate | Mixed angles from several reflectors |
| Thermal load | Concentrated at the rod entrance, which heats and needs cooling | Spread across heads, each with its own cooling |
| Failure signature | Gradual dimming; the whole area falls together | A stripe or a corner goes dark |
| Service | One lamp, one alignment | Many lamps; matched aging, replaced in sets |
Read the last two rows first: they decide how you spend maintenance time. A homogenizer concentrates failure into one lamp and one alignment, simple to reason about and expensive to interrupt. An array distributes failure, which keeps running but produces a gradient easily mistaken for a real sample effect.
The homogenizer loses flux at every internal reflection and again at the entrance aperture, and the loss climbs with rod length. A long rod mixes better and outputs less, so the design sits at the shortest length that reaches the required uniformity — which is why working distance is not a free parameter.
Heat lands in a different place in each design. In a homogenizer the entrance face is the hot spot: ultraviolet and infrared that the rod material absorbs raise its temperature, and the rod then re-radiates into the optical path. A cold mirror or a band filter upstream of the rod removes that load before it becomes a drift problem. In an array the heat is spread across heads, which is easier to cool but adds up in the room, and each head contributes its own infrared to the sample.
Neither route escapes the infrared question: for a temperature-sensitive experiment, the filter belongs upstream of whichever element spreads the beam.
The homogenizer route is maintained like a single-source instrument: one lamp to age, one alignment to check, one entrance face to keep clean. Contamination there degrades uniformity before it reduces output, so the beam dims unevenly and no power reading shows it.
The array route is maintained like a fixture: lamps age at different rates, and a head slightly dimmer than its neighbors tilts the whole pattern. Replace in sets, record which positions changed, and expect a gradient two months after a single replacement.
Both routes need the same discipline about the working plane. Uniformity is specified at a distance, and it degrades on either side of it; a sample sitting a few centimeters closer than the design plane sees a different beam.
Specify four things together or the number is meaningless: area, band, working distance and the spectral range over which the band applies. A small central patch always looks better than the area you actually illuminate.
Acceptance is a grid scan with a calibrated meter at the plane you will use, after the source has reached thermal equilibrium, with window and filter in place. Sample corners and edges as carefully as the center, and finely enough to see the structure: head spacing in an array, or the rod boundary in a homogenizer, sets the grid. Scan twice, cold and warm; a beam that shifts with temperature will fail a long run.
The definition of the uniformity metric, and why max-min and deviation figures disagree, are set out in the large-area uniformity reference; take the metric from there and apply it at your own plane. If the beam must stay centered over hours, check position against a target at the start and end of a run.
It redistributes flux, it does not create it. A rod will smooth an arc image or a reflector structure, but a sector that is blocked or a lamp that is failing stays visible at the output, because the mixing only averages what enters. Fix the input first, then judge the homogenizer.
Yes, in age and in type. Mismatched lamps tilt the pattern rather than dimming it, which is harder to notice and easier to misread as a sample effect. Replace in sets, keep a log of positions and dates, and re-scan the plane after any replacement.
Almost always a difference in plane or in conditioning. The specification is quoted at a design distance with a warm source; a sample closer or farther, a window that was added, or a scan taken during warm-up will all read worse. Reproduce the specified conditions before concluding the source is out of tolerance.
