Micro Array Condenser Design

[Chacob]

I have been working on new condenser ideas and I wanted to share one that I have been working on the past few weeks.

This new condenser is built around the LED Grid Illumination. My thought was that maybe instead of taking this array and placing it really close to the sample (as is the current way to utilize the array), I could instead project the array onto the sample from farther away. This comes with a few benefits.

1. The LED grid can get really hot while in use and quickly evaporate a wet sample
2. Projecting the grid down to a smaller area may allow for more light to hit the sample and create a clearer image at higher NA

To project this grid down onto the sample is much harder to faciliate than I thought it would be. The first thing that needs to happen is I need to collimate all the light coming from each LED on the board. To do that we need a lens for each LED to act as it collimator. When looking for lenses online I found 3mm ball lenses that would work - small catch, they cost 20$ each. To avoid spending over a grand on lenses I opted for the next best thing I could find, cheap boiling glass. I have used these in my testing and they are certainly not as optically clear as the real deal but they do produce roughly collimated light so they’re good enough.

With the boiling glassware I was able to create my first collimation array attempt. The optical train looks like the following:

Now you may notice right away there is a huge problem with this setup, however, I did not put this into the sim before resin printing my first go of it and was scratcing my head wondering why the image in the microscope was so horrible. The RGB LED dies have 3 discrete LEDs on the chip - this is a problem. Because they are seperated by some non-zero distance, so the collimation ends up spreading them apart.

It took me some time to work out, but I’m quite proud of my solution to this problem. Placing the lens farther from the source allows for the distances between the LEDs to be less significant in the generated image - this however does not create collimated light. To then collimate I added a second lens after the first that focuses on the conjugate image of the LEDs. This is a form of kohler illumination! Focusing first lens, defocusing collimating lens. Both the distance and double lens create a much more even collimated output. Here is the optical train for that:

This setup also allows for me to place a diaphram between the two lenses to clean up the output a bit. Also, the output is still a bit messy, but the center collimated light of all 3 colors which is important!

With this new optical train I was able to draft out parts to print:

The design has 5 printed parts total. If I was better at resin printing perhaps I could combine a few, but the support side of the prints always destory features.

The LED array fits into a seperating part with tubes for each LED leading up to the first lens (I did not want an LED to shine into a lens that is not in its train). All parts are held to the condenser via a long m2.5 screw. The scale here is very small.

Here is it placed atop a cheap Abbe condenser I got from China for around 13$:

With it I was able to capture the following at 20x. Pond sample, Stentors and Algae:

My slide and cover slip were very dirty so there are lots of extra abberations

These stentors are more impressive in real time as they move around - you can really see the detail in their moving parts.

I think with some more tweaking this could become a powerful illumination design.

There are a few things that I think this setup would benefit from:

1. Better lenses. I just tossed the “lenses” in there straight from the boiling bag but I think that they have a less than perfect outside finish and lack sphereicity. I plan on making a micro tumbler to tumble them in polish compound to help with that.
2. More precise lens positioning. It is really really hard to bechmark these “lenses” because of how small they are. I am placing them based on the sphere lens equations of focal distance and I have a feeling they are incorrect. And at this scale even a few tenths of a mm can create a worse image.
3. Add a diffuse layer after the LEDs. I have no idea how I will manage it at this scale - but it should help with the color abberations!

This has been fun to work on. If anyone has any suggestions for this project please let me know. I am constantly iterating on it and would to hear if anyone has ideas on how to improve this system.

Cheers!

- Jacob

[r.w.bowman]

This is very cool indeed, nice work!

My immediate thought is: could you use something like this to eliminate the need for the Abbe condenser lens, and potentially improve the working distance?

I’ve mused about putting lens arrays in front of LED arrays for a while, but never done it myself in the lab. Your solution of having a bit more space between the LEDs and the array is a neat one: it nicely gets around the separation of RGB emitters. If you deliberately misalign the two ball lenses relative to each other, you might be able to redirect the beam and add a little focal power - potentially focusing all the lenslets to the same place.

I also wondered if there’s somewhere in this system you could put a diffuser without wasting too much light, which might make the illumination more even - not that it seems like that’s much of a problem.

Finally, the group at Strathclyde University published an OpenFlexure with resin printed lenses: they don’t outperform the glass objective, but I think that might be the ideal technology to make a lens array for the illumination. I should talk to them!

[Chacob]

Thanks!

I think that removing the abbe condenser would be possible if the shperes I used were smaller. The current 3mm spheres nearly max out the available surface area. Playing around in the sim show that considerable overlap is needed to get decent convergence of the beams.

I don’t think there is enough space with these 3mm spheres to achieve that. The lenses in the center would have to be removed and because the lense are in a 2d arrangement they will quickly overlap as they get pressed to the center. Here is a cross section shoing the distances between lenses. (from a side view, showing both array layers of lenses)

I do plan on putting a diffuse layer atop of each LED chip. I’m going to cut out 64 2x2mm polypropylene squares and press fit them into the LED tubes. I will need to make a punch tool for that..

I did get some crystal clear resin, so I may attempt to make a new array once I’m done with this project. It will probably be crazy looking due to the chips having 3 LEDs on board. That will be difficult to model. I don’t know of any good design software to help me develop that.

[WilliamW]

For the diffusers, we have recently been seeing degradation of PP close to high power LEDs. PTFE seems to be a more robust choice, although slightly harder to specify as it is sometimes rather clear. We are still considering the best advice overall. Illumination/Diffuser problems (#448) · Issues · OpenFlexure / openflexure-microscope · GitLab

[Chacob]

This is good to know - I have many sheets of PP on the way so I think I’ll burn through them for testing. I did get some milky acrylic, but it is far too thick to use at this scale. If I had a mill I could shave it down to thickness bu then I wouldn’t know how to cut out 2x2mm squares…

[Chacob]

Small update for one of the items on my todos for this project:

I built a small tumbler to test improving the boiling balls. The goal was to improve the surface finish to hopefully get them to a better optical quality.

This is a conical tube that has a smal motor with a wheel attached that makes contact and spins the tube. I add the boiling spheres to this tube with a slurry of various polishing compounds.

To test the spheres when polishing is complete I built a small optical testing rig. This was built with the components I had laying around - if anyone has any comments on how to improve this I am all ears.

The measuring is done with an LED pointed at a photo resistor. There is a small insert that allows me to design new lens configs to test with this rig. The insert in these images is the following.

I also made a collimating variant of the insert which looks like the following:

To compare the lenses at various states of polish - I start by taking a reading in air (no lenses in the insert) and then use that as a baseline to compare all the lenses. When looking at the following graphics note that 100% is the lensless air measurement. Ultimately we’re really measuring the voltage that is being allowed to reach the analog pin on the esp8266. So a qualitative measurement is the best I can get with these components.

Here is the first run of polishing. Each measurement is the average of 3 lenses at that stage.

This graph was made with measuring with the first insert, which only contains one lens.

The graph is ordered by what I thought would be the least polished to the most polished.

My assumptions were wrong - I thought the cerium oxide would perform much better but that was not the case. I thought that maybe the photo resistor was the problem - perhaps if the more polished lenses were focusing the light too much on one spot the photo resistor would not activate as much due to its design limitations. In response to this data I made a new insert type - the collimating insert (image 2 above). This would avoid focusing the light on the photo resistor and also allow me to multiply the clarity losses of the lenses, which, in theory would allow me to see differences between the lenses more easily with the setup.
Here are the results for the collimating variant of the insert:

Note- I made two version of the collimating insert, this was a good opportunity to test what the real lens distance should be (although the data is not too convincing as to which one is correct).

Note 2- Data is missing for the regular cerium oxide 6 hour and 12 hour because I rinsed them again for the previous graph’s data. I washed all the ones I had saved.

Here we see a similar trend as before, the silicon carbide winning out. Oddly the 12 hour cerium oxide caught up the in the non-optimal collimating distance. I might have to chalk that up to variance for now.

I’m still waiting on the diffuser material to arrive. So in the meantime I plan on exploring a few other polishing times/compunds. I wish to max out the clarity on these boiling balls.

If anyone notices anything wrong with the setup that may be giving bad data please let me know!

Cheers,

Jacob

[Chacob]

Another thought about the cerium oxide vs the silicon carbine is the difficulty in washing it off. I found online that it can be acid washed. I wonder if that would help with the data.