40x test stack

[descall]

Dear Forum,

Below is a test stack (65 images) which shows a bulbil within the leaf axil of the moss Pohlia bulbifera. The first is PMax and the second is DMap output straight from ZS (no cropping, just resizing to 1024 px width and 72 dpi). The set-up is illustrated here. The objective is the Nikon M Plan 40x/0.5 ELWD positioned 200 mm from the camera sensor (by the way, can someone tell me why it is put at 200 mm on bellows rather than 210 mm to achieve 40x?). The DMap surface of the bulbil looks rather nice, I think.




After using a 10x objective I was surprised to see how much more dust shows up at 40x, so I spent quite some time cleaning the sensor and lens before taking the above stack. Even after that, there are still a few imposters. The various settings are as follows:

Camera settings (Canon 50D)
picture style ‘faithful’
jpg image capture
1/200 second exposure
2nd curtain shutter sync
colour temperature set to 3000K
ISO 200

Flash settings (580EX II)
manual mode
1/16+0.7 power

ZS settings
settling time: 2 s
Time after last shutter pulse per focus step: 2 s
step size: 0.002 mm
Prerun distance: 1 mm

I understand from Rik that the blurring at the edges of the images are to do with my camera being mounted out-of-line with StackShot's axis of movement, which I tried to fix but appear to have failed so far. I'll try again.

Any comments on possible improvements would be most welcome. I think I could put a decent image together using the above within ZS and PS, but would like to fix as many problems as possible before starting any retouching. Many thanks for your comments.

By the way, the background is black flocking positioned about 6 cm behind the subject, showing up in the images as dark blue (no doubt there is some interesting explanation for that).

Best wishes,
Des

[ChrisR]

Couple of quezzies Des
what are you doing about a lens hood?
Bearing in mind all the light you're interested in comes from a half millimetre area, there's scope for a lot of unwanted light. Some black paper teepees with silver foil outside - or white or silver tape which holds them together, can help.


And why use 3000K? Flash is 5600 to 6000 usually. That accounts for the blue background. I leave mine on "flash"...


The 10mm you're missing is the distance the eyepiece normally sits away from the aerial image inside the microscope tube - which you aren't using.

[rjlittlefield]

This is promising. It's definitely not easy to work at 40X.

The blotchy background and OOF areas in your DMap result are due to having the slider set too far left. See the tutorial "How To Use DMap" for more info. Basically you want to make OOF areas go black in preview.

--Rik

[Chris S.]

Des, looks like a very decent result for a first test with the 40x ELWD. It’s amazing how well the StackShot rail performs at this magnification—which I suspect is considerably higher than what the designers were concentrating on when they developed it. If I had one (I use the StackShot controller, but not the rail), I’d be terribly curious to see if it could be made even better with a finer lead screw.

You and I have some overlapping interests, as I also shoot quite a few mosses and other cryptogams.

OK, thoughts—some of them pretty picky, especially considering that this was just a test shot. So perhaps useful, perhaps not.

I’d probably choose an increment of about 1.1 microns, rather than two, for this optic and subject. Also, I find myself wishing the stack were a little bit deeper; the out of focus bits are rendered in a very pleasing way (in the PMax), but my brain wants a little bit more to be in focus.

The highlights on the surface of the bulbil are a bit strong for my tastes. Here is where cross polarizing attachments come in handy. I’d use x-pol as a dimmer switch to dial back the highlights, not eliminate them; full x-pol would take much of the surface texture (and optical drama) out of this photograph. Also, the placement of the highlights feels too far forward on the individual bumps—if they were a bit further back, they might take on more apparent curvature, which would work well to convey the shape of the bumps. So you might want to consider moving the flash and/or diffuser, and perhaps masking portions of the diffuser.

[descall]

Thanks for the further comments and advice; much appreciated. I tried another test stack with WB set to 'Flash' and a small lens hood attached to the objective, as illustrated kindly by ChrisR. I used the DMap image, retouching with the PMax image and a small amount of PS. Here is the result:



Matters are improving, but I think I need better lighting and a deeper stack. It is astonishing to be able to see the chloroplasts (the darker green spots) within the plant cells! The above image is not cropped and the smudgy edges have gone since the previous stack (I think this may have been due to my tightening the 'Lock' dial on the bellows, which I may not have done previously).

[ChrisR]

Your persistence is paying off..
As an experiment, a suggestion would be to make sub-stacks of a few of the layers in the picture. The leaf(?) at the front might be an obvious first choice, then the pair of round buds(?) behind.
That way you can clean each part up, and put them in the right order in Photoshop, Gimp etc, as layers. You can mask out, or simply delete, anything you don't want from an upper layer, and a lower one will show instead. It's good for dealing with areas where say, a behind-leaf is crossed by another in front, and the image of the behind one has hazy halos where the front one crosses. It's not quick or simple or perfect though!

[Chris S.]

Des, that looks quite nice. I like how you've incorporated some backlighting that showcases the cellular structure, and also used the light to convey shape and texture. The colors look much better at this color temperature setting. Tame the highlights, and you're golden.

What was your increment for this image?

Regarding x-pol, I've been meaning to put a post together with a parts list for an RMS-thread polarizing setup. I've posted the parts for a Mitutoyo mount, but while that's the same in concept, the bits are different. Unfortunately, I'm going to have to wade through my notes to identify which parts I ended up using in the final version. In general, it amounts to this:

Get some "TECHSPEC® Visible Linear Polarizing Laminated Film" from Edmund Optics and put it in front of your flash (if not using a diffuser) or (if using a diffuser) between diffuser and subject. If diffusing, keep the light source to one side of the subject, or prepare to deal with more complicated issues.

Get a TECHSPEC® Unmounted Linear Glass Polarizer from Edmund Optics (do not get the cheaper ones not labeled "Techspec") and mount it between your objective and cone-shaped adapter.

It's the mounting bits I need to look up. You buy a small threaded tube, put an RMS male thread adapter at one end, an RMS female thread adapter at the other end, and two rings in the middle to hold the filter. Offhand, I suspect I used a Thorlabs SM1 (one-inch) tube and appropriate matching parts.

FYI, I'm in the thinking stage of upgrading this to permit me to drop filters in and rotate them without disturbing the lens. That may involve machining an opening in the tube, reinforcing it externally, and adding a light-blocking cover. It might be based on a Thorlabs SM1 tube, an Edmund Optics C tube, or something larger.

Cheers,

--Chris

[descall]

Many thanks for the further advice and comments. The increment between the images for the above stack was 2 microns (I'll try 1.1 next time).

Chris S - if you manage to dig out your notes on which mounting bits you used for the polarizer I'd be very interested to try this method.

Best wishes, Des

[ChrisR]

I put a sub-stacking example up, because I don't recall seeing much. Might be worth a try. At least you don't have to re-shoot!

[descall]

Below is another attempt at 40x with a different subject. Same settings and set-up as above, but with 1/4 power flash and steeping increments of 1.5 microns (229 images in all). I haven't tried the 'slab stacking' yet, but will give it a go when I've figured that out. The below is based on the DMap output, retouching the edges with the PMax output and then cleaning up the colour balance, etc, in PS. I think this is looking OK?



Some things that are nice to see include:
1. The lighting from above has high-lighted the thickened cell walls of this plant nicely
2. As the light passes through the lower surface of the leaf, several of the papillae (small knob-like outgrowths) on the surface of the leaf cells are illuminated
3. The surface detail of the gemmae (big balls within the leaf) is revealed quite nicely (notice the single gemmae nestled in the channel at the base of the excurrent nerve)
4. The gemmae appear to be a little dehydrated, since the surface walls have collapsed inwards somewhat (I’ve only seen them before at full turgor, mounted in water on a microscope slide, when the surface walls bulge outwards)

Any comments with regards to further improvement would, of course, be much appreciated. If I had shot this at 1.1 micron intervals do you think a noticeable improvement in quality would have resulted (compared to 1.5 microns)? The reason I didn't was that it would have created quite a lot of extra images and there was already 229 with which to deal.

Best wishes,
Des

ps. Forgot to say, this is the moss Syntrichia papillosa (usually grows as an epiphyte on tree trunks)

[Chris S.]

[rjlittlefield]

This looks good. The rule I use about step size is that if I can't see bands of blur, the step size was fine enough. The required step size varies quite a bit depending on subject and image size. Normally with a 40X objective I'd be down around 1-2 microns. The point of that would be to avoid significant banding when viewing at actual pixels with any subject. With this subject I'd be very surprised if you saw any significant difference going from 1.5 to 1.1 microns.

By the way, your initial posting here had me a bit confused. I first picked it up when the post read "and steeping increments of 15 microns (229 images in all)". That struck me as unreasonably large, and I was going to inquire whether the number was correct. I see now that you've edited the post to insert the previously missing decimal points, which makes a lot more sense.

--Rik

[ChrisR]

Des, I think this is the biggest imprrovement I can remember, within so few posts!
If you look at Nikon's views of DOF
http://www.microscopyu.com/tutorials/java/depthoffield/index.html
it seems to suggest you're fine enough. Always worth a try though. In Zerene you can always pick every second, or third, etc, frame directly.
You can do it indirectly by selecting the frames you want of course Try assembing every third frame for half the stack, then all of them for the rest. I tend to see less diffrerence than I imagined, which is probably due to my poor technique - vibrations etc, the lens not being perfect, etc. If you can't see where you switched, you can draw conclusions ...!

[DQE]

Excerpt, from Rik:

"The rule I use about step size is that if I can't see bands of blur, the step size was fine enough. The required step size varies quite a bit depending on subject and image size. Normally with a 40X objective I'd be down around 1-2 microns. The point of that would be to avoid significant banding when viewing at actual pixels with any subject."
----------------------------------
Isn't this a classic sampling rate situation, as analyzed by Nyquist through his famous sampling theorem for band-limited signals?

https://secure.wikimedia.org/wikipedia/en/wiki/Nyquist%E2%80%93Shannon_sampling_theorem

Walking further out onto this "plank", I wonder if it would be possible to estimate the (spatial frequency) bandwidth, in units of cycles/mm, along the depth axis? This way, we'd use standard quantitative units to describe sampling rates.

Wouldn't classic aliasing phenomena occur due to undersampling along the depth axis? A lot depends on how rapidly the optics blurs the scene as a function of distance from the center of the slice. Similarly, Moire phenomena should be visible with undersampled periodic test patterns or scenes extending along the depth axis. Some biological structures of transparent objects are quasi-periodic, too, and subject to Moire.

Another ad-hoc thought: Would it be informative to acquire a microscope slide with a high-frequency periodic test pattern as an demanding but revealing sampling rate "scene"? If this periodic test pattern were a spatial square wave with a frequency of say 500-1000 lines per mm, one could make a stack of the test pattern to investigate the adequacy of the spacing of one's stack.

EDIT: here's a link to a candiate test pattern, available up to 600 line pairs/mm:

http://www.edmundoptics.com/products/displayproduct.cfm?productid=1831

What's confusing me is that I am not used to thinking of blur regions created by undersampling. Perhaps I'm thinking about undersampling when the sampling aperture extends the full distance between sampling points. In stacking, the effective sampling aperture is set by optics and is possibly something roughly like a quasi-Gaussian shape or a quasi-parabola shape, when viewed along the depth axis. Sampling with a finite aperture is different from sampling with a zero-width slice thickness, of course.

------------
A mostly unrelated side note:

During my R&D career, much of which was spent analyzing the imaging characteristics of medical x-ray imaging systems, it was learned that the noise pattern of x-ray film grain (or that of the incident x-ray photons) extends to very high spatial frequencies, and considerable extra effort is required to bandlimit a digitally sampled radiograph. If one does this properly, one can obtain accurate analysis of the image noise in the film and/or the signal content of radiographs.
_________________
-Phil

"Diffraction never sleeps"

[rjlittlefield]