Thank you for the lens position info very helpful.
I tried to move the assembly up with the 10x/0.30 objective, but there wasn’t any additional travel.
So, I did add a 0.35in spacer to the assembly.
The focus worked out and the image was better.
Resolution (r) = λ/(2NA)
based on the first equation: R = 514/(2x0.3) = 857nm = 0.86um
Resolution (r) = 0.61λ/NA
based on the 2nd equation: R = 0.61x514/0.3 = 1045nm = 1.0um
based on the 1951 target, I get [7,6] = 2.19um but that’s the max for my scale!
Looks like I will need a finer scale and a way to move the assembly up further to get rid of the spacer. I started reading about the Delta Stage. It looks like the slot for the objective assembly has a longer slot. Maybe I should move to the Delta stage. I will start printing.
The Delta Stage has got a longer slot, but I don’t think that it is a good solution to your problem. For the imaging that you are doing the Microscope is more appropriate.
It seems that you have objectives with different par-focal lengths. The best thing to do would be to have a different optics module appropriate for that. There are notes on how to generate optics modules with different par-focal distance in the instructions customisations and alternatives. If you make an optics module for 35mm par-focal distance then you can put spacers under the slide if the objective is between 35 and 45mm. That would not affect optical performance.
Thank you for the info.
I assume there is a way to measure the par-focal distance.
These 3 objectives I have all came from the same inverted scope, but it is very old, so probably not the same par-focal distance is very possible.
I assumed they had the same distance, they do work on the original scope. I have to move the objective up quite a bit.
I did try a spacer for the 3.5x/0.09 objective as well, but the image didn’t look so good. The main issue with that objective is the lighting, but there is distortion at the edges.
Looking at the Delta Stage, why isn’t that a better solution. I see that that design may allow for adding an incubator to the scope. Just curious, does it use the same optics?
If you are able to use the objectives on a different microscope, then you just need to focus on something and then measure the distance from the focused object to the shoulder of the lens by the RMS thread. That is the par-focal distance. You say that you need to move the focus between lenses, that is a clear indication that they are not the same par-focal distance. They do sound like rather old lenses, and lens design and manufacture have got a lot better, particularly for giving a wide field of view with flat focus. Are all the objectives marked ‘160/’ somewhere, or are some ‘(infinity)/’?
A 3.5× lens has a very large field of view, compared to the 40-100× that the illumination was designed for. For anything less than about 10× the illumination is not appropriate, neither is the very small step size of the motion.
Some photos of your lenses and microscope would be helpful.
The Delta Stage optics are basically the same as the Microscope, but the camera is rotated 45 degrees relative to the mounting dovetail. Neither version will take much weight on the sample stage, which would make it tricky to mount an incubator on the microscope. Either version can be placed inside an incubator, and the Microscope is smaller so needs smaller incubator, or you get more microscopes in a given incubator. I have not tried the crazy option: having the sample plane completely fixed and hanging the microscope from that by the stage top. Then the microscope moves around below the fixed sample plane. That would probably work, with the microscope stand separated from the main body so that it is not moving the weight of the electronics as well. In that case the fixed sample plane could be a small stage incubator.
I will measure the par-focal distances. I can’t get to the scope until next week.
It may be time to get some newer objectives. But to use the target I have, I will need lower NA objectives or lower magnification. The 3.5x has an NA of .09 so the target I have should work. The .3 NA 10x objective is out of the range of my target.
The other idea is to look for another way to measure the resolution. The targets I have located are just too expensive. The 1951 USAF target I purchased was under $85, but the quality isn’t so good. Maybe I can do something with small diameter beads, there are a number of available fluorescent beads.
I have built a stationary incubator above a 384 microtiter plate, the incubator was kind of large. I will look further into a smaller stationary incubator over say an 8 chamber slide well, maybe a future project!
Reading further about resolution, I think my next test will be with small micrometer beads. I think I have 5 micrometer beads, could also get 1 micrometer beads. At 10x with my 0.22NA objective and a second 10x objective with a 0.33 NA objective. I will image the distance between to beads. When the intensity changes by 20%, that’s the limit of resolution for the system. I will then need to know the number of pixels between the peaks then convert that to micrometers. I then assume I subtract the diameter of one bead. Does this sould like a good next step?
I have a microscope scale that would tell me the number of microns per pixel for a given objective. So I should be able to measure the distance between two beads when the intensity changes 20%. But I do think that you would have to subtract one bead diameter if you use the center of the bead. I haven’t read that, not sure. I guess you could look at the edge of the bead, but how to determine the edge. So I thought the peak dark lines would be the center of the 2 beads.
I did see a bit about using a sharp edge, I will look into that method.
I thing your suggestion to use a sharp edge seems to be a good one. Just image a clean razor blade. Do an intensity profile and measure the distance between 20 and 80% change in intensity. Does that sound correct to anybody? I don’t know yet the significance of the 20 to 80%?
I have measured the par-focal distance of all of my objectives:
3.5x / 0.09NA Bausch & Lomb 33mm
8x / 0.20NA Bausch & Lomb 38mm
10x / 0.30NA Unitron 36mm
10x / 0.22NA Carl Zeiss 44mm
16x / 0.40NA Carl Zeiss 44mm
The scope is an old inverted scope from Carl Zeiss.
I am able to switch between the 2 Carl Zeiss objectives with minimal change in focus, but the other objectives need big changes in focus. They need to be a lot closer.
I did try the spacer under the objective, it seems to give the same quality of image.The overall distance is similar to the (optics_picamera_2_rms_f50d13_parfocal35.stl)
I do have 2um beads, so I will try the resolution test between the beads & measure the distance in pixels. I have a scale for measuring the pixel distance.
I will also try the razor blade edge, but ChatGPT says the bead method is best?
Finally I will try to use red, green and blue LEDs to see the difference in resolution based on wavelength.
I measured the number of microns per pixel with a 600um target for each objective.
Distributed 2um beads on a slide diluted with water, under a cover slip.
I looked at various beads, looking for an increase in intensity between 2 beads. When the intensity increased by 20%, I considered the distance resolved.
Results:
16x/0.4NA expected 0.64 - 0.78um measured 0.274um/pixel resolved to 2.2um
10x/0.22NA expected 1.2 - 1.4um measured 0.430 um/pixel resolved to 3.9um
10x/0.3NA expected 0.86 - 1.0um measured 0.442 um/pixel resolved to 3.1um
8x/0.4NA expected 0.64 - 0.78um measured 0.741 um/pixel resolved to 3.7um
The only thing I would change next time, would be to readjust the camera with each change in objectives. The intensities weren’t consistent between objectives. It was difficult finding enough intensity difference to call the beads resolved.
I tried to do a sharp edge resolution. The test, I think is to look at the change in intensity across the edge. A drop of 80% to 20%, the number of pixels then converted to um. But there is just too much noise in the data to easily pick out the 2 values. It seems to be too subjective?
Final test:
I built up 3 light assemblies with red, green and blue LEDs.
I imaged the 2um beads with each light source after first doing a Full-AUTOCALIBRATE, using the 16x 0.40NA objective.
The resolution for the blue Light (480nm) (0.60 - 0.73um) measured 1.9um (2.2 on white light)
The resolution for the green light (550nm) (0.69 - 0.84um) measured 1.9um but this image didn’t appear as sharp, could have been the focus?
The resolution for the red light (670nm) ((0.84 - 1.0um) measured 2.5um much less resolved
Just to add in possible complications. It is always hard to know if the beads are as opaque in each colour. If they are not they may have less of a hard edge.
Thinking about the razor blade measurement. The high side seems to have a clear maxiumum that could be calculated from and average. You could fit a line to the sloped region to fine the 20 and 80% areas. The issue is that the low side doesn’t seem to have such a clear minimum, there is some sloping. I wonder if this is stray light reflection from the surface. I am not sure what would happen if you put some matt black paint on the blade (though this may affect the sharpness of the edge).
Always the joy of pushing the resolution or accuracy of an instrument is that you make it much more difficult to classify. To me this was the most challenging and fun part of precision instrument design.
I agree, my opinion is that the instrument is doing quite well.
On the bead issue, it might be better for the beads to be black. The beads I am using (so far) are white.
I am going to look into fluorescent beads, maybe they are flat in color. I would also like to look into adding fluorescence, then again look into resolution.
On the sharp edge, the noise in the light area seems to be the issue. I tried to flatten the background, but I still get that noise. Not sure what to do with that method, maybe a curve fit?
My longer range goal is to add super resolution to the scope, but I would need some help to make it low cost. Just reading articles for now.
I redid the bead testing with UV fluorescent beads, 1um.
Using the green LED for illumination, I tested the Fast, Medium and Fine focus, looking again at resolution. Both the Fast and Medium focus gave the same results as previous tests. 1.9um. The theoretical resolution is (0.69 - 0.84um) for the 18x / 0.40 objective.
The fine focus however did give a 1.6um resolution so slightly better in focus.
Next I tried to improve the focus, by just changing the z steps. I tries +100, +200 & -100, -200. At +200 the focus was slightly better, the resolution was 1.37um, so better.
A nice change would be to be able to look at the image zoomed in while you focus. That way you could get a much better focus. That would improve the resolution. Is there a way to do a focal stack now?
On the razor blade method, I tried looking at the distance between the 80%-20% change. Looking at the numbers, the best I could get was 5.71um. Not sure how to improve this method, but the procedure is much more straight forward and probably easier to automate.
What I would do is pick two points that seem between which the red like is very linear. (points 1 and 2 below). And fit a straight line to the points in between with least squares.
I would repeat for the rise (between 3 and 4) and for the top (between 5 and 6)
From here you should be able to calculate the intercept of the red and green line. I would treat this a 0% level. The intercept between the green and blue I would treat as the 100% level.
From these two values you can calculate your 80% and 20% levels, Then using the formula for the green line you should be able to get to sub-pixel distance the position of the 20% and 80% points.
Given a conversions from pixels to um, which you can get from analysing your USAF target you should be able to estimate resolution
Did the linear fits then found the intersections, used the 20% to 80% levels = 1.57um = 1.6um measured resolution.
The theoretical values are ((0.61 - 0.84um) looks good to me.
I like this method, it’s easy, no hunting thru various beads, looking for the threshold.
Is it accurate, can’t say yet, but should be easy to check against wavelengths and tests I am also doing on my Nikon TE300 scope. I have a 20x / 0.45 objective on the Nikon. The images are clearer, but not a big difference. I should be able to see differences.
All this to say thank you, I think I have a base line for further work on super resolution.
