I am starting to build the high resolution system, but need some help on selecting an objective. My goal is to measure the resolution, using a 1951 USAF target. I want to compare the system to my Nikon TE300 system. I would like to try a low cost and a ‘reasonable’ cost objective.
Hi @DougKoebler, it would be really nice to see some different resolution tests.
For objectives, you will want 45mm parfocal distance. You can use 160mm tube length or infinity corrected lenses, but you would need to print the correct optics modules. The standard part in the instructions is for 160mm tube length, the version for infinity corrected lenses is in customisations and alternatives. The infinity optics module is slightly longer, so you also need a taller base.
A Plan Achromatic objective is the base specification that would be appropriate. Non-plan lenses will have too little of the field of view in good focus.
There are so many different objective brands to choose from. Many people on the forum seem to be using Amscope. They have a range of low to mid cost objectives which have clear specifications. For mid-price I have used Motic lenses, but there are many others. You can probably use the lenses from the TE300, to show any differences in the rest of the optical system.
I will look over the Amscope objectives, many to choose from? I see a Motic 40x B series, 160mm with a .45NA for $57. The 40x objective on my Nikon is .60NA, Plan Fluor, ELWD infinity. I think for now better to start with a low end Amscope. Can anyone recommend an Amscope?
I am looking for a low cost 40x objective, plan 160mm tube length, 0.65 NA from AmScope.
AmScope 40X Plan Achromatic Compound Microscope Objective Lens, $56.99
Has anyone used this objective?
These two posts seem to reference the same lens:
I don’t know if they were actually used.
William, I think my best first build is to use the 40x Mitotic lens. I did read and look over all the suggested objectives. My concern is working distance. I will probably look at dividing yeast cells in a liquid, then later look at some live cells. I will measure the resolution then compare it to my Nikon scope. Most of the work I do is at 10 and 20x, so I will test the resolution again. My plan is to image the USAF 1951 target and then measure the resolution using the image.
Thanks so much for the references. I will continue to look for any suggested objectives. Thanks again for your help.
Doug
1 Like
I have finally completed the build of the microscope with “High Resolution Optics”. It took me a while, but I think it is working just fine, thank you to all.
My goal is to measure the resolution using a 1951 USAF target, looking for some help.
I did take an image of the target, now I want to test the resolution.
Is anyone familiar with this target and the procedure for measuring resolution.
I am using a Zeiss 10/0.22 objective.
1 Like
Great!
There are many different definitions of resolution. They all give roughly the same answer, but can differ by maybe two times so you need to pick a particular way of measuring and stick to that. The USAF target essentially pre-dates digital imaging. It has the sets of three black lines at different spacings and the simplest way to use it is to look at the image and decide which set of lines have become blurred together. That is relatively easy to do, but different people will not agree. One person comparing a couple of things will get a reasonable comparison.
For example, in your image you can clearly resolve everything in group 4 and group 5. It gets bad somewhere in group 6. The USAF target specification then tells you what line spacing it is (group 6, element 2 is 7um line width). This is where personal judgement comes in a bit - I might say element 2, you might say element 4 - which is 7um or 5um line width. This is quite poor because of the reduction in image quality in the posted thumbnail image, with the full image you can almost certainly see more. Remember on your microscope that the live image is a preview at a lower pixel count, always use the full saved images for comparisons.
You will need to move the target around and double check the focus, as the resolution and focal position may be different in the centre compared to the edges of the field of view. This is where your expensive Nikon ELWD is likely to be significantly better than a cheaper lens. For lenses with NA above about 0.4 you also need to make sure that you are not limited by the illumination. In principle it is the NA not the magnification that determines the resolution, and to use the full NA of the lens you need at least that NA in from the illumination.
We did write a python script for calibration using a USAF target:
Edited to say that this hasn’t been updated for aaages… and so the setup file seems to require python 3.7. It should probably work on later python version if you remove the line in setup.py that blocks it.
1 Like
My thought is to first get the best image I can, then use ImageJ to look at the intensity of the dark to light transition. When the intensity between the lines is some value ? above the dark lines then that’s the limit of resolution.
I suppose I could invert the image so the lines are light and the intensity between the lines is darker would be better. Then look for the acceptable value like 75% for example. But what is the max intensity, 255 and what is the dip say 255(.75) = 191 as an example.
Does that make sense?
I will look up the accepted Rayleigh limit for a microscope.
I really appreciate your response, but it’s over my head (for now). I will give it a try, once I figure out what to do, It looks very interesting.
Here is an example of the procedure I am thinking about. I took the image of the USAF target, inverted the image then did an intensity profile. The bright pixel is 104, the dark is 43.
I realize I need to get the best image, but with this image this target seems to represent my best resolution. Does that sound correct?
2 Likes
That would be a good procedure. Make sure the actual maximum value us not 255. You need to be sure that the image is not saturated.
Is your example an actual image from your microscope? The maximum resolution there is one pixel per dark or light line, which is unusual. We would expect the lines to be several pixels wide at the point where the optical resolution blurrs the lines.
I agree that is the last resolved element. The USAF target is designed so that it can be read by eye. It is nice to have an algorithm to be onjective. But I think that the algorithm is that from top to bottom of the 3 bars there should be 3 peaks that all stand above the 2 troughs. By that metric you can just about select the group below if you only look at the horizontal bars, but the vertical bars are not resolved.
In this specific case as it is 1px per line as @William has said, this resolution limit could be set by how well the bars line up with the pixels.
On the topic of our program micat, I agree it isn’t the friendliest thing to install and use at the moment. It would be great to get this into the microscope software as a plug-in. The last version I used (in 2019!!) doesn’t actually give out resolution, instead it is using the largest groups in the image to calculate the field of view and um per px. These are from its outputs with a 10x objective:
However, it also provides an output showing you how it matched each group:
Here the blue bars show the detected edges. The red horizontal line is a y axis and the vertical lines are the intensity in the centre of the bars. The green line then represents a cross correlation as the detected central bar as it is swept over the image.
I think this code could be adjusted to find smaller groups and attempt the same thing to detect resolution.
@JohemianKnapsody Has there been any thought about a micat plugin for calibration? How are you doing the resolution checks for the HQ camera?
The example above image is not the high resolution image, but it is from the microscope, it’s 200x156 pixels.
The high resolution image is 3280x2464 pixels, RGB. So I converted it to grayscale then inverted the image. I then used ImageJ Plot Profile function to get the Gray Value across the target. The max is 39, min 28 so 28/39 = 72%. The next smaller target is max 30, min 28 so 28/30 = 93% that doesn’t look resolved (to me).
1 Like
According to my data on the target, I can resolve to [7,3].
I looked up the width of the line, a chart in Wikipedia, [7,3] = 3.10micrometers.
If I calculate the resolution of my objective: using green light, my NA = 0.22 for the 10x objective
(I don’t know the condenser NA.)
R = 1.22(514nm)/0.22 for my objective I get 2.85micrometers or 2,850nm
Is this correct or do I have more research to do, trying to understand?
1 Like
Disclaimer - I am not the optics guy in the OFM team!
Theoretical Resolution
Looking at MicroscopyU, they give 3 possible resolution calculations
r = \frac{\lambda}{2\,\text{NA}}
r = \frac{0.61\lambda}{\text{NA}}
r = \frac{1.22\lambda}{\text{NA}_\text{obj} + \text{NA}_\text{cond}}
The are all approximations with different assumptions. Based on the first two, for your NA = 0.22, \lambda = 514 nm. I calculate resolution as 1.17 um and 1.43 um.
In the case of the 3rd equation, I once measured the NA of the condenser as 0.48, this would then give us 0.90 um resolution. However, it doesn’t make sense to me that we can almost arbitrarily make the resolution higher for a low NA objective with a really high NA condenser. I would guess that the 3rd equation assumes the NA of the condenser is lower than the objective and in this case the condenser NA limits the location. So I am going to ignore this number. (@WilliamW can let me know if I am misunderstanding something.)
From this we are saying that in a best case scenario we would expect a resolution of about 1.2-1.5um. Assuming perfect focus, perfect lighting, infinitely sharp target, no image compression, and pure 514nm green light.
Measured Resolution
I agree that you have resolved 7-3 and not 7-4. 7-3 has a line width of 3.1um, 7-4 has a line width of 2.76um.
The exact meaning of resolution in microscopy terms is based on point spread functions, that should form an airy disk pattern. This is effectively a measure of how the lens blurs the image.
Looking at the ThorLabs explanation of how to use a USAF target:
“The largest set of non-distinguishable horizontal and vertical lines determines the resolving power of the imaging system.”
So based on this measure, the resolution would be 2.76um.
Comparison
Measured Resolution = 2.8-3.1um
Theoretical maximum resolution = 1.2-1.5um
So it looks like the system isn’t quite reaching the theoretical maximum resolution, but then I think that it is quite normal to not hit the theoretical maximum.
This may be limited by the illumination, or by the focus (what happens if you take a series of images with small z-distances in between). Also we have “warm” light with quite a bit of red in it, for the red light we expect a resolution of closer to 2.1um (@750nm). I am not sure if calculating your greyscale image from just the green channel would help at all?
The other thing is you are probably using the JPEG image? This will have compression artifacts. It is possible to get the raw image from the camera sensor (Bayer data). This may help you squeeze out a tiny bit more resolution (for quite a bit of effort!)
All in all I think it shows that your microscope is working pretty well, and now you are in the realm of trying to squeeze out maximum performance, which always gets very complex!
I agree with you that the microscope is doing very good.
Thank you so much on your analysis. I too am not an optics guy, but for some reason I am very interested.My background is physics and mechanical engineering, retired now. I am interested in working towards super resolution in the future.
I will read up on the NA of a condenser.
I have a Nikon TE300 inverted microscope, the condenser NA is 0.52.
My next idea is to first try using a green LED to see if there is any difference, then blue but probably not, the RGB values were all close but there are differences.
I will also try a couple of other objectives I have first, 10x/0.30 (Unitron) and 3.5x/0.09 (Bausch & Lomb). Both of these objectives need to be much closer to focus, maybe 1 or 2cm. I think I will need to make spacers for them. Do you know the allowable distance for the Z travel? My thought is to just add a spacer to each objective so there isn’t much difference in the Z motion.
Most objectives have a parfocal distance of 45mm. That is the distance from the objective shoulder where the screw thread ends, up to the sample. If you do need to adjust for a different parfocal distance I would move the position of the mounting screw/nut. If you put a spacer between the objective and the tube lens then it will not be working at its designed working length and this will cause aberrations.
The way we calculate the lens positions can be found here:
https://build.openflexure.org/openflexure-microscope/v7.0.0-beta2/info_pages/imaging_optics_explanation.html
1 Like
Thank you, I will give that a try.
1 Like