Building a Microscope from Camera Lenses

Microscopes are powerful tools which allow you to see extremely small features, down to less than a thousand of a millimetre. While microscopes can be affordable with prices starting at 100 EUR, good quality material can cost you several thousands. Differences will stand mainly in resolution and aberrations in the image. With low price microscopes, you will not be able to resolve very small details and your image may be blurry at the borders or you may have the different colours of the sample not mixing really well. High quality microscopes require high quality optics and fine adjustments which have their price.

In this post, I would like to show how it is possible to build a good quality microscope from inexpensive camera lenses. Please have a look at Figure 1 for an example of image obtained with the setup presented here. It is a sample of Saccharomyces cerevisiae yeasts (better known as Baker’s yeasts). The yeasts are about 3-8 µm large and the total field of view of the image is about 420 µm x 340 µm. On the right handside of the image, you can see a larger object which looks to me like some living creature with a flagella but I cannot identify it (so if you have any idea of what this is, please send me an e-mail as it was in the water I use daily in my experiments!).

A conventional, transmission bright field microscope, consists of three important parts:

- The imaging system which consists of a microscopy objective and a tube lens.

- The illumination system which produces an homogeneous illumination for the sample.

- Some accurate positioning system to move the sample and hold the various sub-systems together.

The quality of the image will be limited by the quality of the optics used with a lot of stress in the microscopy objective (the illumination system also play an important role here). Good quality microscopes require using “plan achromat” objectives which mean that the objective is achromatic (it handles polychromatic light so the different colours will mix correctly in the final image) and that the image of a plan will be a plan (the image will not blur at the borders). The resolution of the microscope will be limited by diffractive effects which are linked to the numerical aperture of the lens which is the size of the cone of light that is accepted by the objective from the sample. High resolution requires high numerical aperture and large angular acceptance cone. To give you an order of magnitude, the ultimate limit for traditional microscopy is on the order of a fifth of a micrometer and it requires very complex setup (high angular acceptance objectives, thin layer of oil between the sample and the objective and usage of light close to the UV range). The microscope presented here can reveal details on the order of 1 µm. You can see a close-up of budding cells on Figure 2. The small bud on the bottom of the image is about 1.65 µm large and appears relatively clearly. This is to be compared with [»] former experiments using the same objectives and which shown resolutions down to 0.95 µm which seems to be confirmed here.

Obviously, if you spend a lot of money into high quality objectives, you don’t want to ruin everything by using a cheap tube lens, so it is important that the tube lens does not introduce aberrations into the image such as chromatism. The role of the tube lens is to achieve magnification. If you use a 6 mm focal length objective and a 200 mm tube lens, the overall magnification will be 200/6 ≈ 35x. This may seems small (especially when you see microscopes advertising a “1500x” magnification) but it is much enough to image the sample on a camera sensor. If you have pixels of 5.5 µm, each pixel will then represents details of about 0.165 µm. But don’t forget that we are limited by the resolution of the objective and so every detail smaller than 1 µm here will be blurred. So you can see that a 35x magnification is actually already too high because we are into the empty magnification range: we increase the size of the image but we do not get any more details. Empty magnification is something you do not want because as you increase object sizes, you also decrease the total field of view. And it is always better to have a large field of view so you can look at more objects on the same image.

In the setup here, I am using a NAVTAR 6mm f/1.4 megapixel c-mount objective as microscopy objective (you can read more about it [»] here) and a Nikkor AI 80-200 mm f/4 camera lens as the tube lens. You can find the Nikkor lens for less than 150 EUR in second-hand shops. It is a relatively good lens and has the advantage here that we can use the zoom of the lens to modify the magnification ratio anywhere between 13.5x and 33.5x. When using these lenses, it is important to set the f-stop number to its maximum value to open the lens and to adjust the focus of the tube lens to infinity (this is another advantage of the Nikkor lens as it has an hard-stop at infinity; modern lenses do not have this anymore). F-stop is another way of talking about the numerical aperture and is more popular in photography while the numerical aperture is more popular in optical sciences. You can read more about it in [»] this previous post.

Also, I told you the resolution is limited by the spread of the light cone reaching the objective which put some emphasis on proper illumination of the sample. While the numerical aperture is fixed by how the objective is actually built, it is possible to artificially lower the spreading of the light cone by using restricted illumination. This is called the realized numerical aperture. So you can have the best microscopy objective in the world, if you use poor illumination, you will not get more details in your objects. Also, a good illumination setup allows you to have a uniform background which is free from glares. It is recommended to use high numerical aperture aspheric lenses for the illumination setup but I will come back on this later.

The setup used here is shown on Figure 3. From right to left you can see the light source (a bright blue LED), an aspheric lens to catch as much light as possible from the LED source, a diffuser screen (you can also use an iris), a second aspheric lens to create a proper illumination scheme, the sample mounted on a XYZ translation stage, the NAVTAR 6mm f/1.4 objective, a mirror to reflect the light to 90° (because my breadboard was too small to put everything in line – that’s just it) and the Nikkor lens with the CMOS camera attached at the opposite end. Here I am using a c-mount scientific CMOS camera, but you can as well use your DSLR camera to have a larger field of view as they usually have much more pixels than the small c-mount cameras.

As you can see on Figure 3, the mounting is quite impressive and represent about 75% of the cost of the setup. Rigid mounting is really important when taking picture at such small resolutions but also it is very important to be able to accurately move the sample along the XYZ axis. This is because the depth of field at such numerical aperture is very shallow – only a few micrometers thick. Moving off the sample by as little as a hundred of a millimetre can make your image completely blurry and thus worthless. This is where camera lenses become handy because you can finely tune the focus by using the focus ring of the objective. An economy XY translation stage from Thorlabs will then do the lateral movement of the sample. Note that it should be possible to make the design much simpler if we forget about XY motion and that we assemble everything from cheaper supports. I will leave that for later improvements!

Let us now dig into more complex stuff: aligning the setup with proper illumination scheme. Have a check on Figure 4 for a schematic of the optical principle and compare it with the setup of Figure 3 as you read the instructions.

First, place the light source anywhere you want, at some height so we don’t have problem with the large Nikkor lens after. Place the first condenser lens (Thorlabs ACL2520U – aspheric condenser lens, 20 mm focal length, 0.60 NA) and the iris such that you make an image of the light source on the centre of the iris. Place the second condenser lens at the focal length of the iris such that it produces light rays at infinity (you can use the Nikkor lens at infinity and the camera to align the condenser lens – simply move the condenser lens away from the iris until you see a clear image of the iris on your computer). You can then place the sample in the beam path and the NAVTAR objective at the focal distance of the sample (once again, you can use your Nikkor lens at infinity to do that). Do the fine adjustments with the focus ring of the NAVTAR objective until the sample is in focus (nb: while this used to work in the past, I could not get it working this time and so I had to mount the objective onto a translation stage… I will investigate on this because this was a very nice feature with other lenses). Then, place the Nikkor lens at some distance from the NAVTAR objective. You should not be able to see the image of the LED; if it is the case, move back the Nikkor lens until you get an uniform background.

This alignment is a simplified version of the Köhler illumination scheme. If you follow the source light path on Figure 4, you will see that no image of the light source is formed on the detector plane. Instead, each point of the sample is illuminated by different points of the light source which allows then having a uniform background. The role of the iris is to control the light cone of the illumination and so the realized numerical aperture of the setup. It works as an image-telecentric setup which selects more or less the light rays that will exit parallel to the optical axis. Indeed, when the iris is completely closed (mind experiment of a single point pinhole), only the light rays that exit parallel will remain because geometrical optics tell us that rays passing by the focal point exits parallel to the optical axis. As we open the iris, we will allow more oblique rays to exit the system which creates then a larger realized numerical aperture for the objective. In this regard, it is important to se high numerical aperture condensers because the overall realized numerical aperture will be limited by the numerical aperture of the condensers and it is important to use aspheric lens to achieve a good focusing of light rays on the iris (otherelse the spherical aberrations will modify this directional selectivity).

Here, as I wanted to maximize the realized numerical aperture, I have followed recommendations of Thorlabs and used a diffusive grit. But I have also tested the “official” iris version and it works fine in this setup; so you can use it.

So, no more excuses to study biology!