Dodt gradient contrast
Re: Dodt gradient contrast
Huh, yeah, the original word in the patent actually is Lichtschwert and German Wikipedia has no "disambiguation" page so.. Hans-Ulrich Dodt is a fan?
Re: Dodt gradient contrast
A few details from the patent regarding the design of the mask, with DeepL translations:
Radial width of the annulus:
Radial width of the annulus:
Die Breite des wirksamen Viertelsringes (8), der noch zur Beleuchtung in der Apertur des Objektes wirksam ist, beträgt vorteilhaft 1/2 bis 1/10 des Radius der Objektpupille des Mikroskops. Mit schmaler werdendem Beleuchtungsring steigt der Kontrast des Bildes.
Angular extent of the annulus:The width of the effective quarter ring (8), which is still effective for illumination in the aperture of the object, is advantageously 1/2 to 1/10 of the radius of the object pupil of the microscope. As the illumination ring becomes narrower, the contrast of the image increases.
Es können auch freie Sektoren bis zu 180 Grad verwendet werden, wobei 90 Grad-Sektoren einen maximalen Kontrast in den beiden orthogonalen Raumrichtungen geben.
Radial position of the annulus, I think:Free sectors up to 180 degrees can also be used, with 90 degree sectors giving maximum contrast in the two orthogonal spatial directions.
Die Blende ist so bemessen, daß auch noch der Teil des Quadranten in der Kondensorapertur, der nicht mehr in der Objektpupille abgebildet wird, beleuchtet wird.
(I am interpreting the last one as saying the illuminated area is intended to straddle the edge of the objective aperture radially?)The aperture is dimensioned in such a way that the part of the quadrant in the condenser aperture which is no longer imaged in the object pupil is also illuminated.
Re: Dodt gradient contrast
This is how I was thinking the Thorlabs system corresponds to the Dodt patent. The patent and the Thorlabs Dodt diagram are showing different sets of rays -- focusing in the objective real focal plane and conjugates in the patent diagram, focusing in the sample plane and conjugates in the Thorlabs Dodt diagram -- but I think they are consistent. The Thorlabs condenser diagram shows both sets.hans wrote: ↑Fri Dec 18, 2020 1:46 amIn the Thorlabs diagram it is not clear how the mask is positioned relative to an image of the light source. However it does show the mask being in a region where the field stop is collimated, which I think is not necessarily inconsistent with it being in a Fourier plane conjugate to the objective rear aperture as in Dodt contrast.microbear wrote: ↑Thu Dec 17, 2020 8:08 pmThorlabs seems to use a different system, interestingly! At least based on what they show on the website, their system is no Dodt contrast. They replaced the two lenses with their engineered diffuser, if I understand it correctly. The engineered diffuser is not the Dodt diffuser but a second one before the annulus. After the annulus it seems they have a traditional diffuser.
- Attachments
-
- dodt-correspondence.png (164.02 KiB) Viewed 3069 times
-
BrianBurnes
- Posts: 37
- Joined: Sat Feb 08, 2020 5:05 pm
Re: Dodt gradient contrast
I happen to be a native German speaker and have read the patent for a bit, and I think you are mostly correct in your investigations hans. The Thorlabs and the patent diagrams specify the same optical system, and simply show complementary planes.
Patentspeak is rather dense, but some salient parts that stood out to me:
- The main advantage of Dodt contrast over oblique illumination lies in how it masks out part of the illumination beam. Both oblique and Dodt illumination can be viewed as masking out part of the Fourier spectrum of the illuminating light beam. Oblique illumination represents a hard filter that sets most Fourier coefficients to zero; this manipulation of the spectrum removes low frequency components and leads to increased contrast, but the hard nature of the filter leads to ringing artifacts and over-expressed contrast. On the other hand, Dodt contrast is closer to what would be called a windowing filter in digital signal processing, in that it masks out Fourier components but has a soft edge (the "gradient") to prevent faux ringing artifacts.
- Using a mask and diffuser is actually not critical. These two optical elements happen to be a convenient way of creating a spatial filter with a feathered edge by creating a hard mask and blurring it with a diffuser, but other optical elements could be used. The patent suggests alternatives, such as a glass slide partially deposition coated with a reflective material, or a computer manipulated liquid crystal element.
Following on the previous point, I don't think the optical setup needs to be quite as complex. Lenses L1 and L2 are there to allow for manipulating the light in the condenser aperture plane at an arbitrary point in the light path, without having to change the rest of the optical setup: The condenser can be left undisturbed, and the light source simply needs to be moved back a little (which is easily done in a research optical system). However, I think that if you are willing to modify your condenser, then you could simply place the light modulator directly into the condensor aperture plane, where you would normally place your DF filters. Similarly, you don't necessarily need a mask+diffusor: If you can manufacture a spatial mask with a feathered edge, you should get the same effect. Since we are not dealing with imaging thick brain tissue with IR as Dr Dodt was, this should be very easy to DIY - printed acetate may be a good starting point, though it will likely have too much light leakage.
Also, because this was the source of some entertainment: "Lichtschwert" translates literally to "light sword" and can have varying meanings: In SciFi it of course refers to the light saber; in lighting design, it refers to elements that "cut" the light path, in the form of a light shelf; and in optics it can refer to "sword shaped" light, i.e. a beam of light that narrows to a point.
Patentspeak is rather dense, but some salient parts that stood out to me:
- The main advantage of Dodt contrast over oblique illumination lies in how it masks out part of the illumination beam. Both oblique and Dodt illumination can be viewed as masking out part of the Fourier spectrum of the illuminating light beam. Oblique illumination represents a hard filter that sets most Fourier coefficients to zero; this manipulation of the spectrum removes low frequency components and leads to increased contrast, but the hard nature of the filter leads to ringing artifacts and over-expressed contrast. On the other hand, Dodt contrast is closer to what would be called a windowing filter in digital signal processing, in that it masks out Fourier components but has a soft edge (the "gradient") to prevent faux ringing artifacts.
- Using a mask and diffuser is actually not critical. These two optical elements happen to be a convenient way of creating a spatial filter with a feathered edge by creating a hard mask and blurring it with a diffuser, but other optical elements could be used. The patent suggests alternatives, such as a glass slide partially deposition coated with a reflective material, or a computer manipulated liquid crystal element.
Following on the previous point, I don't think the optical setup needs to be quite as complex. Lenses L1 and L2 are there to allow for manipulating the light in the condenser aperture plane at an arbitrary point in the light path, without having to change the rest of the optical setup: The condenser can be left undisturbed, and the light source simply needs to be moved back a little (which is easily done in a research optical system). However, I think that if you are willing to modify your condenser, then you could simply place the light modulator directly into the condensor aperture plane, where you would normally place your DF filters. Similarly, you don't necessarily need a mask+diffusor: If you can manufacture a spatial mask with a feathered edge, you should get the same effect. Since we are not dealing with imaging thick brain tissue with IR as Dr Dodt was, this should be very easy to DIY - printed acetate may be a good starting point, though it will likely have too much light leakage.
Also, because this was the source of some entertainment: "Lichtschwert" translates literally to "light sword" and can have varying meanings: In SciFi it of course refers to the light saber; in lighting design, it refers to elements that "cut" the light path, in the form of a light shelf; and in optics it can refer to "sword shaped" light, i.e. a beam of light that narrows to a point.
Re: Dodt gradient contrast
Thanks Brian, there are a couple other things I am still not sure about, maybe you got a better understanding from the patent:
- Typical strength of blurring: is the quarter annulus still supposed to be fairly well defined, or more like a smooth gradient across the aperture?
- Exact relation between radius of objective aperture and projected radius of quarter annulus: I saw the bit mentioned a couple posts ago which I was interpreting as saying there should be some illumination outside the objective aperture which would normally be blocked with no specimen present, anything more precise you saw?
- Attachments
-
- R060-L090-W020-B000.png (42.75 KiB) Viewed 3030 times
-
- R060-L090-W020-B020.png (62.23 KiB) Viewed 3030 times
-
- R060-L090-W020-B040.png (70.51 KiB) Viewed 3030 times
-
- R060-L090-W020-B080.png (85.92 KiB) Viewed 3030 times
-
BrianBurnes
- Posts: 37
- Joined: Sat Feb 08, 2020 5:05 pm
Re: Dodt gradient contrast
It helps to not just skim the last page of a thread - I see now that all of my previous points have already been made. At least it's good that we all arrived at the same conclusion 
On a second read of the patent I also noticed the last paragraph of section 1 which states "the spatial filter for gradient contrast can also be inserted into the rear focal plane of the condensor, or in any other such conjugated plane in the illumination path", so a simpler setup without the refocusing lenses L1/L2 should work.
As for the first question: The patent gives no clear answer. It states that "no part of the Fourier plane stays completely dark, so that all Fourier components contribute to image formation", i.e. there should be some transmission everywhere, which is a given with acetate. In section 2 it mentions that diffusors of different strengths could be used and at different distances to the mask, so the blurriness is highly variable. My best guess (coming from Fourier filtering) is that low blurriness comes with more contrast but less resolution and more ringing, and vice versa for higher blurriness.
As for the second: It does say that parts of the condenser aperture are illuminated that lie outside the cone that can be imaged by the objective. It adds that "because an immersion condenser with higher N.A. than the objective is used, additional contrast- and resolution enhancing effects of DF microscopy come into play".
Some other notes: The patent specifies that lens L2 can be moved with the intent to change the slope of the gradient, as well as the projected size of the annulus in a convenient way. If the masks can be made cheaply and exchanged easily I don't think the movable lens is critical. Another interesting point made in the patent is that for thick sections, part of the contrast improvement lies in the fact that regions above and below the focal plane receive less light and out-of-focus scattering is reduced, so the design of the mask is not just about improving Fourier components.
On a second read of the patent I also noticed the last paragraph of section 1 which states "the spatial filter for gradient contrast can also be inserted into the rear focal plane of the condensor, or in any other such conjugated plane in the illumination path", so a simpler setup without the refocusing lenses L1/L2 should work.As for the first question: The patent gives no clear answer. It states that "no part of the Fourier plane stays completely dark, so that all Fourier components contribute to image formation", i.e. there should be some transmission everywhere, which is a given with acetate. In section 2 it mentions that diffusors of different strengths could be used and at different distances to the mask, so the blurriness is highly variable. My best guess (coming from Fourier filtering) is that low blurriness comes with more contrast but less resolution and more ringing, and vice versa for higher blurriness.
As for the second: It does say that parts of the condenser aperture are illuminated that lie outside the cone that can be imaged by the objective. It adds that "because an immersion condenser with higher N.A. than the objective is used, additional contrast- and resolution enhancing effects of DF microscopy come into play".
Some other notes: The patent specifies that lens L2 can be moved with the intent to change the slope of the gradient, as well as the projected size of the annulus in a convenient way. If the masks can be made cheaply and exchanged easily I don't think the movable lens is critical. Another interesting point made in the patent is that for thick sections, part of the contrast improvement lies in the fact that regions above and below the focal plane receive less light and out-of-focus scattering is reduced, so the design of the mask is not just about improving Fourier components.
Re: Dodt gradient contrast
Regarding the strength of blurring, perhaps another hint is the "curved, tapered lightsaber" description, which could be taken to imply that the blurring should not completely obscure the curvature of the annulus as it does in the third and fourth examples in my previous post?
Laser printing is giving disappointing results so far in terms of toner density but I am thinking it looks like some sort of intentional image preprocessing rather than an actual hardware limitation. Anything I print (the generated masks, printer's own internal demo page, various printer test images from the internet) comes out looking like some fairly strong highpass/edge-enhancing/unsharp mask sort of processing was applied and none of the settings I have messed with (media type, toner density, resolution modes "ProRes 1200" vs. "FastRes 1200" vs. "600 DPI", enable/disable "resolution enhancement technology") affect it. Looking at the results under a microscope I can see the toner coverage is close to 100% within ~1 mm of a sharp edge, so the hardware seems capable, but then the interior of solid black regions fades to a much lower density. I assume this is something done to try to give a similar perceptual effect while reducing toner usage? Or is there actually some physical limitation of the laser printing process that make it impossible to maintain high toner density across large areas? Has anyone else run into this while trying to print illumination masks? The printer is HP M401dne.
Laser printing is giving disappointing results so far in terms of toner density but I am thinking it looks like some sort of intentional image preprocessing rather than an actual hardware limitation. Anything I print (the generated masks, printer's own internal demo page, various printer test images from the internet) comes out looking like some fairly strong highpass/edge-enhancing/unsharp mask sort of processing was applied and none of the settings I have messed with (media type, toner density, resolution modes "ProRes 1200" vs. "FastRes 1200" vs. "600 DPI", enable/disable "resolution enhancement technology") affect it. Looking at the results under a microscope I can see the toner coverage is close to 100% within ~1 mm of a sharp edge, so the hardware seems capable, but then the interior of solid black regions fades to a much lower density. I assume this is something done to try to give a similar perceptual effect while reducing toner usage? Or is there actually some physical limitation of the laser printing process that make it impossible to maintain high toner density across large areas? Has anyone else run into this while trying to print illumination masks? The printer is HP M401dne.