By the razor blade method, here are my results:
8x 0.20NA (1.4-1.7um theoretical) 2.22um measured
10x 0.22NA (1.25 - 1.53um theoretical) 1.96um measured
10x 0.30NA (0.93 - 1.12um theoretical) 1.6um measured
16x 0.40NA (0.69 - 0.84um theoretical) 1.1um measured
By the bead method with 1um white beads
8x 0.20NA (no results)
10x 0.22NA 3.9um
10x 0.30NA 3.1um
16x 0.40NA 2.2um
16x 0.40NA 1.9um (blue LED)
16x 0.40NA 1.9um (green LED)
16x 0.40NA 2.47um (red LED)
16x 0.40NA 1.37um (manual focus -200 from FAST, green LED)
16x 0.40NA 1.9um (FAST focus, green LED)
16x 0.40NA 1.9um (MEDIUM focus, green LED)
16x 0.40NA 1.6um FINE focus, green LED)
Based on the 1951 USAF target, with the 10x 0.30NA objective I reached the limit of the scale 2.19um.
Without purchasing a more expensive target, I would like to find a way to verify the accuracy of these tests. My opinion, the results are very good, they do correlate, but which one is more accurate?
The bead method is closest to the 1951 target, the razor blade method is a much simpler test. One other possibility to make both methods closer would be to use a wider range, say 10% to 90% for the razor blade method?
My plan is to look for some papers on these methods, looking for accuracy.
1 Like
Awesome work.
I think there are enough slightly different definitions of resolutions that the most important thing is to be clear and consistent when making comparisons rather than trying to get âexactâ numbers.
Maybe so, but I am interested in super resolution longer term. I would like to know if Iâm on the right track. I think both the bead method and the razor blade methods will tell me if I improve the resolution. But how much is my issue. I am hopeful that someone has worked that out, I just need to find the appropriate publication. It seems to me that fluorescent beads are preferred for measuring resolution, but that doesnât work with visible imaging.
1 Like
Well I think you need to have something like fluorescent beads to do super resolution. Because you need some sort of information in an imaging mode that can be average down to localise something to a resolution than cannot normally be resolved. It isnât possible to image something like a blade with super resolution.
Better late than never, here I am!
Iâve turned micat into a Thing for v3 (equivalent of an extension on v2), after Tanzania Iâll test it and get it merged. I think it probably wants to be a wizard, similar to the first time use thing
Weâve got a final year student working on testing expensive USAF target (group 9) vs printed edge vs razors edge for automated calibration, will report back
1 Like
Sorry I have been away so long, visiting my son in Bern.
Please keep me updated, very interested.
1 Like
I have another interest in resolution.
I would like to explore how the lighting effects resolution.
Resolution (r) = 1.22λ/(NA(obj) + NA(cond) from MicroscopyU
Is anyone familiar with the NA of the current condenser, how itâs calculated?
I am looking at building 2 light sources, one with a Cree Green LED with a 30deg viewing angle and a second with a Cree Green LED with a 15deg viewing angle.
Will one of these change / improve the resolution of the system?
One experiment I can try is to remove the visible condenser from my scope (Nikon TE300) and use it on the OpenFlexure system, to see if I get a higher resolution. That will take some effort, but as usual, I am interested.
My understanding is that you want to try to match the NA of the objective and the condenser, and then get the condenser focused as well as possible.
The list time I did a measurement of the condenser NA I did it with the old condenser with the 5mm LED. It was done roughly with graph paper and a clamp stand:
If you move The module up from an arbitrary (marked) position measure the diameter of the spot of light on the graph paper. You can then extrapolate out the angle and hence the NA. Last time we did it we got a value of 0.483, from a very crude statistical model the uncertainty was ±0.004, but no attempt was made to actually properly address systematics. I think that 0.48±0.02 is probably more reasonable, it was too long ago to do a better job on the uncertainty.
Also note my ±0.02, is not saying all condensers are within 0.02 of 0.48, just that I am 68% sure that on this day, this specific one, was between 0.46 and .50, and 95% sure it was between 0.44 and 0.52. We havenât done any measurement of the variance between different condensers. This is probably much less now we have the illumination board and the fixed diffuser, but I am not sure how consistent the PMMA lenses are.
3 Likes
I will order the LEDs should start testing next week.
Here is an interesting video on visible light / NA effects on resolution. I havenât yet watched all his videos, but very interesting.
I built up a green LED with 30deg dispersion and a second LED with 15deg dispersion. I measured the NA as suggested above. I really didnât see any difference between the 2 LEDs.
I do see that the 15deg LED is a bit brighter, but no real difference.
In both cases, I did a Full Auto Calibrate with each LED. I didnât see any difference in the resolution.
I did see some white areas in the images that I canât seem to find in the hardware, but it does seem to come from the lighting not the objective.
I tried cleaning the lens, but it might be on the inside surface, still looking. I have the lighting set where the the front optics assembly is flush with the back arm. Is there a procedure for setting the light assembly at a preferred height?
Sorry my error, I didnât remove the slide when doing the Full Auto Calibrate! The images now look very good, with both LEDs.
The images for both LEDs look the same, with the 15degree dispersion slightly brighter.
Same question above, is there a correct height for the light and does that have any effect on the resolution?
2 Likes
For the LEDs, as long as the angle is big enough that the light fills the condenser lens it will not make a difference. Brightness will be less for larger angles as you lose light around the lens.
The height does matter. In the diagram for NA measurement in @j.stirling 's post 48 in this thread there is a focal point for the condenser. This is not only the brightest point, but also it is the only point where there is light at all angles across all of the illuminated area. Below the focus, light at large angles is only towards the edges and at any point only at a large angle from one side.
The plastic lenses are not great lenses and do not have a really well defined focus. There is a larger vertical region where the spot is small-ish. Within that region it should not make much difference, but really pushing resolution on a high NA objective there may be a sweet spot.
1 Like
I have been testing the resolution, razor blade method, of the Green LEDs, 30 and 15deg dispersion. I do see a difference in resolution:
10x 0.22NA Green LED, 30deg dispersion 2.38um
10x 0.22NA Green LED, 15deg dispersion 1.18um
10x 0.22NA White LED, 30deg dispersion 2.07um
The differences are small, the 15deg does give a brighter image but not much. The height of the light source doesnât seem to change the resolution much.
These numbers are all better than my earlier tests with the 10x 0.22NA objective (3.9um). I wonder if the difference has to do with focus on the razor blade, or the type of razor blade. In the previous tests I used a new X-ACTO blade. This time I used a standard razor blade. I will probably look at changes in focus, to see if I get much variation with resolution.
3 Likes
I tested the effect of focus on resolution using the knife edge method with a 10x / 0.22NA objective. I used fast focus then moved the objective to +5000 steps then back to 2000, 1000, 500, 400, 300, 200, 0, -100, -200, -1000, -2000 steps. I got the best resolution numbers between +200 and +500 steps.
| Steps |
UM Resolution |
| -2000 |
8.63 |
| -1000 |
5.53 |
| -200 |
3.19 |
| -100 |
2.9 |
| 0 |
2.24 |
| 100 |
2 |
| 200 |
1.58 |
| 300 |
2.37 |
| 400 |
1.89 |
| 500 |
1.63 |
| 1000 |
2.99 |
| 2000 |
5.55 |
|
|
| The expected value is r = 0.61 lambda/NA =0.61(.520um)/0.22 = 1.44um |
|
The expected value is r = 0.61 lambda/NA =0.61(.520um)/0.22 = 1.44um
1 Like
