4
$\begingroup$

I'd like to design a device that takes 4 channels of light from sets of LEDs and multiplexes them into one optical path, which finally creates a rectangle pattern of light that is approximately 30mm tall and 1mm wide. This is to be used with a line scan camera. In practice, only one of the 4 channels would be illuminated at any given time.

Here's a rough drawing of what I'm imagining:

sketch of a multiplexing setup

I've thought about doing it with mirrors as shown below, but obviously my LEDs are not perfectly collimated sources of light so this would not work. Perhaps this can be solved with a pair of convex cylindrical lenses facing each other to roughly collimate the light, but that would require a lot of lenses and money.

enter image description here

I've also seen a lab setup in this paper, but once again my light sources are not collimated and I suspect a lot of light would be lost to the beam splitters. The LEDs I'm using are in the 100s of lumens.

multiplexing setup from a paper

I would appreciate if anyone could point me in the right direction on this design problem. I'm familiar with the basics of optics, but clearly I'm not an expert. Thank you.

Improve this question
$\endgroup$
6
  • $\begingroup$ How many of those lumens are you willing to give up? $\endgroup$
    – Jon Custer
    Jun 10 '21 at 1:42
  • $\begingroup$ Ideally as few as possible. Let's say 50% as a conservative estimate? I have no clue what a reasonable number is here. I could always add more LEDs to make up for the losses. $\endgroup$
    – John V.
    Jun 10 '21 at 1:58
  • 1
    $\begingroup$ Ok. Illuminate a 30x1mm slit as evenly as possible, now us a 1:1 optical relay with a reasonably good quality large lens to image the slit onto your device. Play with mag/demag of the optical relay if you want. Try to keep the angles illuminating the slit reasonable to avoid losing too much in the relay. $\endgroup$
    – Jon Custer
    Jun 10 '21 at 2:15
  • $\begingroup$ Do you need the beams to be coaxial, or just focussed to the same location? $\endgroup$
    – Carl Witthoft
    Jun 10 '21 at 12:27
  • $\begingroup$ Focused to the same location, but the light has to be as even as possible. The light would be illuminating a target that is scanned by a line scan camera, so if there are large variations in exposure along the image then quality will be unacceptable. $\endgroup$
    – John V.
    Jun 11 '21 at 5:19
5
$\begingroup$

This is going to be really hard to do with LEDs since they're not collimated. If they were instead lasers then you would have some options at your disposal.

  • If you have control over the polarization then you can combine the beams using polarizing beam splitters with lower power loss (but you would have differently polarized light at the output, may or may not be a problem)
  • You can use non-polarizing beamsplitters as you've shown. You would lose 50% of the power of each beam
  • You can use an acousto optic modulator or deflector. This is a device designed for spatial multiplexing of light. I'll leave it up to you to look it up. However, I'm not sure it will work over such a wide range of optical wavelengths.
  • Probably the optimal way to do this would be dichroics. These are optics that reflect light above (or below) a certain wavelength and transmit light below (or above) a certain wavelength.
Improve this answer
$\endgroup$
2
  • 1
    $\begingroup$ +1 for mentioning the dichroic mirrors. $\endgroup$
    – sleepy
    Jun 10 '21 at 7:17
  • $\begingroup$ I agree that dichroics are the most interesting option here. I will take a look, thank you very much for your suggestion. I would like to make it a solid-state solution (i.e. not having to move the LEDs or a mirror) so this is a promising path forward. $\endgroup$
    – John V.
    Jun 11 '21 at 5:17
2
$\begingroup$

To do this efficiently, you want an optical device known as an X-cube. This is a cube that has three input faces: for red, green, blue. The output face combines all of those. Look up "X-cube optics" on the web, (so you don't get X-cube toys).

Since you also want a fourth channel, that is a complication. People would typically add a dichroic plate beamsplitter in the beam path. The fourth wavelength is injected into this beamsplitter. (The main beam passes straight through).

It is unusual to combine more than 3 colors. It may require specialized coatings that will cost money.

Your output beam will never be perfectly collimated. The LEDs have a finite spatial width at the source. When you use a lens to collimate light from a finite size object, the output beam can never be perfectly focused. (This is true even ignoring diffraction, which is another source of beam spreading).

Improve this answer
$\endgroup$
2
  • $\begingroup$ This is a great idea, thank you for the suggestion. I think the X-cube could work, even with 4 channels. In theory I could put two pairs of LEDs per side so I'd only need two sides that way. I'm looking at using a TIR + Fresnel lens combination (like in this paper) so even a rough focus and collimation should be adequate. My only concern would be the power loss in the X-cube, I will need to do some research to see what kinds of losses I could expect. Once again thank you very much. $\endgroup$
    – John V.
    Jun 11 '21 at 5:24
  • $\begingroup$ My 2x2 setup would not work since each face of the X-cube (dichroic cube) only works for one color. Back to the drawing board :) $\endgroup$
    – John V.
    Jun 11 '21 at 5:33
0
$\begingroup$

It sounds like you want a projector. Normally you use it to project a multicolored image on a screen. You want to project 4 different uniformly colored images on a small screen, one at a time.

Or perhaps you just want to put 4 different LEDs some distance from the screen and turn on one at a time.

Improve this answer
$\endgroup$
1
  • $\begingroup$ My goal is to project light in a specific pattern, not an actual image. 4 LEDs some distance from the target would work, but the problem is that the working distance is relatively short (i.e. <300mm) and the LEDs cannot be placed close enough to eliminate alignment issues. $\endgroup$
    – John V.
    Jun 11 '21 at 5:16

Not the answer you're looking for? Browse other questions tagged or ask your own question.