0.72 NA Condenser for darkfield at 40x and below

[Chacob]

Hi all,

I recently set out to make a custom condenser for my v7 scope. I started using the scope only a few weeks ago and am really impressed with the image quality from the standard condenser that comes with the base kit. However, I also bought the components to build the LED array and I fitted that to the scope - and that took the visuals to a whole new level. Here is a little guy I found in some pond water illuminated by the LED array(10x,LED darkfield, LEDs set to a high green ratio):

The LED comes with some drawbacks in the form of many visual artifacts for objects not in the currently focused depth of field. This is caused by the LEDs sitting close to the optic with many discrete locations of each LED, each LED can create its own artifact (with 64 LEDs that noise can add up quickly). This is especially noticeable on slides with many objects (at various depths) or slides viewed at higher magnifications.

My theoretical solution to this LED artifact problem is to condense the LED array to a high NA. I have found the prospect of doing that a bit daunting so I’m going to make a more traditional condenser first.

Here is a look at my optical test bench:

This is where I had fun relearning everything I had forgotten from physics class. This paired with this online simulator allowed me to build out my designs. This also allowed me to easily measure the focal distances of the generic lenses I bought on amazon.

This testing led to the first iteration of my condenser:

Unfortunately I did not know what NA was so I built probably the worst darkfield condenser on the planet with this design -resulting in a measured NA of 0.14! After educating myself on the topic some more I have V2:

This condenser can achieve a theoretical 0.72 NA (perhaps a bit more with the spherical aberration caused by the lenses). The lenses from Amazon cost about 0.80 cents each, and require 3 lenses per condenser. Uses a handful of m2.5 screws, and the stock LED board that came with my v7 kit. So it is quite cheap to build each unit.

The aperture is fully 3D printed (minus the screws used to hold it onto the condenser). It can be found here: https://www.thingiverse.com/thing:6584051

The light stops are little 3D printed plates that slide into a slot just after the aperture.

I finished assembly today and did some quick tests with a dry slide laying around that has a piece of duckweed/dust on it. I have honed in on the light block size needed for my NA for 10x and 40x. Here is how that looks.

First is the duckween in 10x with both brightfield and darkfield. Note that I have not spent any time honing in the focal distance of the condenser - which I think is causing the brightfield to appear more blown out than it should:

And finally for my 40x test (which I was very nervous about):

In the first of these two 40x images I was using a light stop that was too small and allowed for some light to leak, causing the gray background. There are severe aberrations within these 40x images but the video display is much cleaner so some of it may be due to the pi camera capture settings. (and probably a lot to due with the spherical lenses…)

Things I believe will improve the usefulness of this design:

1. Aspherical lenses to correct the aberrations
2. A purpose build aperture - this one can create odd edges

Going forward I wish to attempt the following.

1. Figure out how to make a micro lens array to collimate the LED array.
2. Develop a method to make my own aspherical lenses with silicon molds
3. Combine 1 and 2 to build a condenser that does not require an aperture nor light stops as those will be controlled by the LED array in code.

If anyone has any interest in how this is constructed I can set aside some time to build out a thingiverse/bambu post with a bill of materials so others can have cheap (somewhat usable) darkfield.

Please let me know if there are any improvements I can make or designs I should look into - I basically started learning about this a week ago so my breadth of knowledge in this domain is really small.

Cheers, Jake

[WilliamW]

Hi @Chacob this is really nice. I am amazed that you fitted it all in and mounted on the normal illumination dovetail rail. The printed iris aperture that you are using is a great addition.

I would certainly like to see the optical train that you are using. Is it effectively a Kohler illumination system?

[Chacob]

Hi @WilliamW thank you for your kind words!

My system is Kohler-like; it is missing a few elements that would be included in that system. I do not have a field diaphragm nor a field lens. Honestly, I did not really understand the Kohler system until you asked me if mine is like it (I had to research it just now and my understanding is still probably lacking). I think because the LED delivers an even field of light, the system is effectively a Kohler system, with the exception of the iris being placed without much care. I just slapped that in there to make everything small.

Here is the optical train (not to scale):

The source is the LED that came with my kit; the collector lens is a spherical plano-convex with a focal length of 30mm and a diameter of 26mm. There is the 3D-printed iris, then a slot for light stops, and finally the stacked condenser lenses, one of which is the same as the first lens but reversed, and on top of that is a 22mm focal length, 20mm diameter lens.

Here is a copy of that optical train in the simulator. There, you can move around the iris or light stop. This also shows the degree at which the spherical aberration affects the train.

[j.stirling]

Woah! This is super cool!

We would love to be able to include some upgraded illumination options in the core project.

[r.w.bowman]

This is really cool. I think a microlens array for the condenser could be a really nice use case for some of the 3D printed lens technology being developed by e.g. groups at Strathclyde University. I really need to get round to trying to make that happen! Putting a lenslet array in front of an LED array potentially gets around the problem you’re having of the discrete array of spots, so if there’s really a way to make them cheaply it would be awesome.

We’ve now got a resin printer to play with - we just need to find someone with the time to play with it!

While the LED array approach is cool, I think a more traditional condenser is often a better option, provided the lenses are good enough (and “good enough” for a condenser isn’t a super high bar). It would be lovely to figure out how to get something like this into the main repo as an upgrade option or something in the future.

[Chacob]

Thanks @j.stirling and @r.w.bowman !

@r.w.bowman I do think the LED Array condenser is out of scope for the tools that I currently have. I did consider getting myself a resin printer over this last week - but decided I would first try to figure out this basic condenser before getting in too deep. I do think it would be possible to create my own FDM printed aspheric lenses. I have cooked up a workflow that I think will work, but I also have access to a metal lathe to assist me which I understand many do not.

I would love to be able to contribute something new to the project if it would be useful!

I’m currently in the process of refining this current design. As I use it there are some obvious design flaws. My todo list for this is to address the following:

• The headspace is artifically low due to the last lens in the train being secured by wrapped plastic on the outside - which loses us about 2mm of room atop the slide. The lens is already press fit into the body so there is no need for that extra bulk or the screws. Easy fix.
• I have reused the dovetail rail to secure the condenser. I am currenty using a crude interference fit which has a nice sturdy feel but is way too awkward to adjust along the height axis because of how much force must be applied to move it up and down. I think this would be easy to resolve with a solution similar to your current condenser design with the lock screw, but another problem that compounds with this problem is that the condenser is not fully in-line with the objective. It pitches down a few degrees which allows for noticable light leakage at higher magnifications. This might be due to the plastic not being rigid enough or shrinkage in the prints. To account for this I think I’ll need to design a whole new illumination mount that will allow for precise pitch/yaw control of the mounted condenser. Harder fix.
• The LED/LED wire connection is currently exposed. That should be housed to not leak light everywhere. Easy fix.
• The design currently uses 3 lense, 2 of the same and 1 that is slightly smaller. The smaller lens offers no real benefit as its NA almost exactly that of the larger lens (the only reason it was added was to allow it to fit past the press fit of the larger lens). This would assist in cutting down cost to obtain parts and help with spheric aberrations. Easy fix.
• The Iris I am using is one that works to manage NA/exposure, but it is quite coarse to adjust. The lever to adjust it has a very small analog range - it also cuts into the edges when fully open which is lowering the total NA. Medium fix.
• Once all of the previous items are finished. Create a set if printable light stops for various NAs that are honed in well specifically for the lenses that are called for in the design. Easy fix.

@r.w.bowman @j.stirling @WilliamW I’d be interested in hearing what requirements you all have to make this something that is viable as a new illumination option for the project (in addition to what I have on my current TODO listed above)? It would be a honor for me to be able to contribute something like this.

Thanks for reading all of that - here are a few frames from a video I took last night of some ciliate eating something - maybe and algea? 40x with the V2 condenser.

[WilliamW]

@Chacob

I’d be interested in hearing what requirements you all have to make this something that is viable as a new illumination option for the project

The starting point is really just something that works, preferably where we understand what parameters matter. This gives us the understanding of what is possible and how to achieve it, also what actually fits in the space.

The next consideration is availability of parts - to be really useful we need to be able to specify clearly what the lenses are, and they need to be reliably available in a wide range of locations. This is particularly tricky for cheaper lenses: the condenser that we specify is a weak point of the existing bill of materials. This does not need to be a consideration yet, if you have found what works it is not a big process to resize for slightly different lenses.

Details of the actual CAD, and instructions, are relatively straightforward when we know what the components are, where they need to be, and what shape holder works.

[j.stirling]

What @William said. Also worth mentioning that there is a lot lower bar for something to be listed in “customisations and alternatives” than to get full build instructions in the main docs. Which requires a lot more testing

[Chacob]

I have spent the past few days working on the design and I believe it is in a decent enough position to share. I made some major changes in both parts accessability and function:

• Removed the Iris. This design does not have a field lens so the iris only cause the NA to be decreased. The only possible benefit would be to cut off the spheric aberration at the edges, but that can be done with a simple light stop.
• Removed the need for all M2.5 hardware - each piece now screws together with printed threads.
• Raised the dovetail so the last lens sits ~6mm above the stage. The focal distance of the lenses is about 6.5mm from the last lens. This allows for the condenser to be pressed all the way down and allows for the light to be focused on the same plane similarly every time.

Next are some benchmarking photos at my 3 main magnifications. Note that these are snips of the preview window so they’re not full resolution. Also, I’m a novice when it comes to getting the camera calibration honed in - I found that when I was testing the 10x looked very good but then another day it looked quite poor because I had tweaked some camera setting and could not figure out how to get it back.

I’ll start with the upper limit of this design; 60x in air:

The darkfield has severe aberrations and the background is partially lit. However, the subject is obvious so that is a minor victory. Perhaps better camera settings would help with clarity and background color.

Next is my favorite and the target of this scope; 40x in air:

With 40x I was able to get some fantastic looking images. Here are 3 darkfield and 1 bright. You’ll notice one of the darkfields has a lit background - that was corrected in the most recent design with a lens tweak.

Finally 10x:

10x works great - which is expected given the 40x looks great. At this magnification I am able to achieve true darkfield. I don’t have a 20x objective to test with but I am guessing that would also look great based on the 40x data.

To build this condenser you need the following purchased components:

• OFM LED board
• 4 self tapping screws that are used with the base condenser
• The spacer and diffuser used in the base condenser
• 3 32mmx8mm lense bought here (only come in pack of 10) PATIKIL Diameter 32mm Height 8mm Flat Convex Lens, 10 Pcs Acrylic LED Optical Convex Condenser Lens for LED Flashlight Bike Head Lamp, Transparent - Amazon.com

You will also need one of each of the following printed components for the body:

LED holder.stl (768.1 KB)

LED cap.stl (146.4 KB)

Double Lens cap.stl (444.4 KB)

Condenser Body.stl (738.0 KB)

Here are some light stops that I messed with to achieve the photos above:

6mm stop.stl (91.5 KB)

24mm stop.stl (98.5 KB)

22mm Stop.stl (97.9 KB)

20mm stop.stl (97.9 KB)

10mm stop.stl (95.0 KB)

8mm stop.stl (93.8 KB)

(I used the 22mm for 40x and the 8/10mm for 10x)

Cross section of the assembly:

Two of the purchased lense press fit into the “Double lens cap” and the last lens press fits into the condenser body towards the top. The LED board/spacer/diffuser screws into the top of the LED holder. Similar to how it is configured within the base condenser.

I think I have pushed these generic amazon lenses about as far as they can go. I have lenses in route to me from China that claim to have decent stats and are also affordable. I plan to use these to see if a new version of this can be made to allow for sharper images to be captured.

If anyone has questions or recommendations on the design please let me know. This has been a fun project to work on in my free time.

[Chacob]

I attempted some kohler-like illumination while I wait for the new lense to arrive.

I used the cheap lenses I had on hand which resulted in a pretty long optical train to get the correct placement of the lenses/apertures.

I quickly ran some algae under it to see how this type of illumination compares to the condenser I just made:

I’ll refer to the last posts condenser as the darkfield condenser and this new one as the kohler condenser.

Here is 40x:

First is the darkfield condenser (one without a light stop so we have bright field comparison)

And now here is the new kohler condenser:

Next is 10x:

First the darkfield condenser (one without the light stop)

And now the new kohler design:

I think the kohler design does achieves a more even background and has pretty crisp details when compared to the brightfield of the ‘darkfield condenser’. It also has a much larger depth of field, which is to be expected. I think this design is nicer to work with when compared to the darkfield condenser for brightfield. I do think if you want the best image quality then the high NA darkfield condenser wins out, but you would need to z-stack to get decent depth of field for brightfield.

That is all for today. Next I hope to work with the new ashperic lenses and will hopefully have some nice comparison images of how those look compared to these plano convex amazon lenses.

[WilliamW]

There is a much larger difference than I would have expected in the brightfield. The Kohler looks awesome.

[ffesti]

I can imagine both condensers might benefit from a better (read “flexure based”) adjustment mechanism. Probably both in z and in xy.

[j.stirling]

100% agree. I think there is a lot of benefit in us using the upright optics mount to create and adjustment for the condenser, especially as we get towards 100x.

x/y adjustment is a bit more complex to make nice and compact, but we do need something better than just adjusting manually at the base of the dovetail!

[WilliamW]

I am not sure xy is necessary, at least until there is a field stop in full Kohler. Even then, it is the field stop that needs xy, not necessarily the whole thing. The Upright has no slots on the upper mount, it always seems to be aligned well just relying on print accuracy. I have used 20× to 100×.

I would like to test removing the slots from the normal illumination dovetail.

[ffesti]

I am thinking about something like +/- 1mm in each direction, may be a bit more in z. So I agree with @WilliamW that the precision of the print is already pretty good. We basically just need to adjust what might not quite match up. Funny enough the slots for moving the dovetail rail around increases the need for a convenient an precise adjustment and its range.

[Chacob]

During my testing I often need to recenter the illumination on the optic. I have been printing with the same offset for each new condenser design but I have a feeling the act of pressing the new condenser into the dovetail is probably knocking it off center. Making a new dovetail with fine xy adjustment screws is on my todo list. I’m not sure if everyone would need this, I would like it for quick testing - a fixed/rigid dovetail that is made to sit correctly may be the best for most people.

[j.stirling]

Would you be able to make technical drawings of any of these parts and to give us permission to redraw them in OpenSCAD? Would be really cool to play with some of this and get it into the main open repo.

Technical drawings aren’t 100% needed but they tend to make it easier to understand the design than simple looking at STLs. Same of any details of calculations for part separation so we can make it parametric.

[ffesti]

May be have the dovetail have a z adjustment and may be also tilt in two directions and make the xy adjustment part of the condenser. I am thinking angled blade flexures wrapping around at 45° and being attached at the front.

[Chacob]

Something to note: I measured the focal lengths for the lenses that I used but when I applied the lens equations to them to find the conjugate image planes (required for the kohler condenser aperture positions) I found that the real position of the aperture was off by quite a large margin. The lenses I used were less than perfect and were the culprit. I’m bringing this up because unless you build out an optical train test bench, like the one pictured in my original post, you won’t be able to determine the position of the elements in a given condenser programatically by the stats of the lenses. The parameters of the new CAD will need to set the relative positions of each object in the train using the real world measurement of the test-benched objects.

@j.stirling I can put these into drawings no problem. I did create both condensers within OnShape so I could also link that instead if you would like to have access to all dimensions/objects. Would you like the source CAD, drawings, or both?

[Chacob]

I think in a perfect world I would want the dovetail to have all degrees of freedom built-in, that way the condensers that we print/test would only have the responsibility of illumination and not also positional adjustment. This would result in less printed material for an array of condensers.