 |
 | Home | Contact Us | FAQ | Search & Site Map | Link to Us |
 | Sign In | Join | Other 45 Sites in Network |
Natural Science Forum / Physics / Optics / July 2005DSLR for photomic - grossly oversampled
I'm Using an EOS 10D mounted on the front photoport of a Zeiss ICM 405. As I have posted before, this setup has FOV issues caused by the small sensor (15.1mm x 22.7mm) compared to the 35mm film the ICM 405 is designed for; I live with this and do more mosaicing on specimens that need it - but it would be nice to have a fix. One solution I considered was to buy a DSLR with a 35mm sensor (1D or 1DS) but this is VERY expensive option and (worse) because the pixels on this sensor are EVEN smaller than those on the 10D, it exacerbates another issue I've been hitting. Namely; small pixels result in gross over-sampling - which is the intended topic of my post... I'm interested to hear others comments and/or experiences on this. Cheers Beats PS. Background information and musings... With 3.5x mag through the photoport and 7.4um pixels on the CMOS sensor of my 10D the following table shows how many pixels cover the smallest resolvable feature for each objective. Objective | Mag N.A. Res Img Pixels -------------+------------------------------------- Neofluar Ph3 |100.0 1.30 0.21 74.04 10.0 Multi Imm | 40.0 0.90 0.31 42.78 5.8 Neofluar | 40.0 0.75 0.37 51.33 6.9 Neofluar Ph2 | 40.0 0.65 0.42 59.23 8.0 Neofluar | 16.0 0.35 0.79 44.00 5.9 Neofluar Ph2 | 16.0 0.35 0.79 44.00 5.9 Planapo | 10.0 0.40 0.69 24.06 3.3 Neofluar | 10.0 0.30 0.92 32.08 4.3 Neofluar | 6.5 0.20 1.38 31.28 4.2 Plan | 2.5 0.10 2.75 24.06 3.3 -------------+------------------------------------- Mag = Magnification of objective N.A. = Numerical aperture of objective Res = Theoretical smallest resolvable unit Img = Size of a unit on sensor (after magnification) Pixels = # sensor pixels covering one resolvable unit OK, Nyquist states you need two samples to digitize a signal, so 2.0 in the "Pixels" column would be ideal. But as you can see, my setup is over-sampling by a factor of 1.5 to 5.0 depending on the objective used. This state of affairs wastes pixels, download time (from the camera) and processing power (on resulting images). Possible solution #1 -------------------- Bigger camera pixels. New camera needed, therefore expensive and maybe not one available with the ~20um pixels needed anyway. Not attractive. Possible solution #2 -------------------- Find a gizmo to reduce image scale (e.g. a 0.5x focal reducer). Tried to find one to solve the FOV problem and failed - even after getting hold of a knowledgeable tech at Zeiss Germany. Maybe I missed a product somewhere? Possible solution #3 -------------------- Use 2x2 or 3x3 binned modes. Yep - works a treat except for one thing. You can't get RAW pixels in binned modes from the 10D, so dynamic range is reduced from 12 bits to 8, which is insufficient for the deconvolution I often apply to my images. I AM using this fix for some images, but it isn't good enough for my most critical pics where the 12 bit dynamic range is essential. Possible solution #4 -------------------- Get higher NA objectives. I would, but I don't think they're available and the cost would be prohibitive anyway. Possible solution #5 -------------------- Go to afocal projection through a trinocular setup. I replaced the binocular head with a trinocular to try this, and didn't like it! The camera is up high and nowhere near as mechanically stable as on the front photoport. I also can't see the camera LCD very well(for controlling exposure times etc). Most importantly, I want to use the trinocular port for a video system, so current geometry is much preferred anyway. Possible solution #6 -------------------- I have no possible solution #6... >I'm Using an EOS 10D mounted on the front photoport of a Zeiss ICM >405. As I have posted before, this setup has FOV issues caused by the >small sensor (15.1mm x 22.7mm) compared to the 35mm film the ICM 405 is >designed for; I live with this and do more mosaicing on specimens that >need it - but it would be nice to have a fix. I'm not familiar with the ICM (?); I use an Ultraphot 3 for most of my photomicroscopy. >One solution I considered was to buy a DSLR with a 35mm sensor (1D or >1DS) Note the 1D does not have a full-frame 35mm sized sensor - it has a 1.3x factor rather than the 1.6x of the 10D/20D/300D/350D series though. > but this is VERY expensive option and (worse) because the pixels >on this sensor are EVEN smaller than those on the 10D, it exacerbates >another issue I've been hitting. Namely; small pixels result in gross >over-sampling - which is the intended topic of my post... One thought - if you could find a second hand 1Ds (Mk 1 version) with an 11 MP full-frame sensor, you could save a lot of money and reduce the oversampling. >I'm interested to hear others comments and/or experiences on this. > [quoted text clipped - 32 lines] >used. This state of affairs wastes pixels, download time (from the >camera) and processing power (on resulting images). You would know your own needs best, of course - but is the oversampling ^that^ much of a problem? You can surely just downsample the files before further processing if necessary. >Possible solution #1 >-------------------- >Bigger camera pixels. New camera needed, therefore expensive and maybe >not one available with the ~20um pixels needed anyway. Not attractive. One of the MF digital backs (does your ICM take such a thing). A second-hand one with lower pixel count should suit. Still very expensive though. >Possible solution #2 >-------------------- >Find a gizmo to reduce image scale (e.g. a 0.5x focal reducer). Tried >to find one to solve the FOV problem and failed - even after getting >hold of a knowledgeable tech at Zeiss Germany. Maybe I missed a product >somewhere? Again, as I am not familiar wit the ICM, this may not suit; however, there is a wide-field 5-position changer which has a 0.63x magnification factor. My specimen does not have a part number visible, but from the catalogue it appears to be 47 31 55. 0.63x is of course 1/1.587, so it should almost exactly cancel out the 1.6x factor of the sensor. It would however not allow rapid switches between the 0.63x and normal (1x) FoV. >Possible solution #3 >-------------------- [quoted text clipped - 9 lines] >Get higher NA objectives. I would, but I don't think they're >available and the cost would be prohibitive anyway. Assuming your stand is a 160mm tube length one, the 63x NA 1.4 is the only one which springs to mind. I paid about £500 for mine. Wouldn't make that much difference, though it ^is^ a superb lens. >Possible solution #5 >-------------------- [quoted text clipped - 9 lines] >-------------------- >I have no possible solution #6... (6) Try using a fixed-lens digicam? In addition to my 10D, I also use a Coolpix 995 fitted directly to a Zeiss KPl 10x20 eyepiece (screwcut to fit) which slides directly into an eyepiece tube. I'm sure something like this could be adapted to fit, though I don't know the configuration of your photoport. The 3 MP would probably be sufficient, and using the zoom control would give you variable FoV. You also get real-time focussing output from the camera. The later version (4500) has 4 MP; later Coolpix models have much wider lens fronts and are harder (but not impossible) to fit. (7) Get a dedicated photomicroscopy camera - see Gary Gaugler's recent post. (8) Try a webcam and lots of stacking. Hope these thoughts help - if your ICM is a modern infinity system most of them may be irrelevant. If we can clarify further, some details of the scope would help. David  SignatureDavid Littlewood I would not say this. It is very likely that you are just using the wrong relay lens for your setup. Take one that is suitable for your scope. What scope are you using anyway? - I am using a D70 with 6MP on a Nikon Eclipse scope. The D70 has a 1.5x cropping factor compared to the "regular" 35mm SLR format. Can you see over-sampling in the following JPEG output: http://www.pbase.com/overney/image/39648499/original.jpg Before I would buy a 35mm sized DSLR for a scope, I would spend the money on a dedicated camera with CCD cooling. Nothing beats this in quality, ease of use and low noise. A DSLR on a scope makes only sense if you use it also for other types of photography. Otherwise, it's a waste of money and much too tedious to make it work. Compared to a Coolpix 5000, the quality is marginally better. Compared to a Coolpix 990, it is better. Compared to a SBIG cooled CCD (even with just with 2 MPixels), the DSLR is worse for low light application. However, the DSLR is better when used with a 4x APO lens with an NA of 0.20 than any dedicated camera with less than 4MPixels but with a Bayer filter. Gregor > I'm Using an EOS 10D mounted on the front photoport of a Zeiss ICM > 405. As I have posted before, this setup has FOV issues caused by the [quoted text clipped - 84 lines] > -------------------- > I have no possible solution #6... > Can you see over-sampling in the following JPEG output: Well, no - but unless it's a crop out of a larger image, this pic isn't at the full resolution of your D70 - it's only around 1500 by 1100 pixels. Isn't the D70 capable of something around 3k by 2k? Clarification - I am not using a relay lens (other than any that might be built into the internals of the ICM405). I'm projecting directly onto the sensor using the front photoport which is actually designed to project directly onto 35mm film frames within a dedicated camera back instead. I have used adaptors to mount the EOS 10D there instead of the (M35?) Zeiss film camera that came with my system. I use the Zeiss exposure control box to hold the internal shutter open in bulb mode - I then use the EOS shutter (with mirror lock) to control exposure times directly. What I need (but can't find) is a "relay lens" offering magnification of 0.6x, but these only seem available for video systems and their tiny little CCD sensors. I can't find one for my setup... I have a coolpix 4500 too - I think the EOS 10D produces better pictures, and offers better control etc. I also have a megapixel cooled CCD I use for astrophotography. I tried this for imaging on the microscope first (before purchasing the 10D). It works, but has even more oversampling as the pixels are 6um compared to the 7.4um of the 10D. FOV is pitiful. You're right though - this is unsurpassed for v. low light work... Anyway - back to the point, while I agree your image doesn't look oversampled, I also can't argue with the numbers I calculated and results I am getting (which agree with the numbers). Do you have more info on your image? What objective, magnification, NA etc? What is the image scale (i.e. how many pixels/micron in the final image). Cheers Beats The picture is a crop to be able to compare it with an image the Coolpix 990 produced. The D70 captures the entire real image with some vignetting. By removing vignetting, I loose roughly 1MPixel of data and hence end up with a 5 MPixel shot. The following image shows the entire FOV but it is then resized to match a practical format for pBase (see Pine Single Needle Leave c.s. at http://www.pbase.com/overney/image/44874329.jpg.) I found that dealing with vignetting is much less painful than trying to find a 0.6x relay lens that covers adequately the entire CCD of a DSLR, which has a 1.5x cropping factor. I haven't found one yet either. (BTW, my scope does no require a relay lens to fully correct for CDM.) I also do not yet believe Nikon's claim that the market will never accept 35mm image sensors in DSLRs ("because they would always remain too expensive and required better corrected lenses, which also add to the cost of ownership"). So far, Nikon decided to stick with the so called DX format, while Canon seems to dance on both parties simultaneously. - Why invest now a huge amount of money into the perfect setup when we might be able to buy a 35mm image sensor, which allows us to bin neighboring pixels to avoid "oversampling" and increase S/N performance? The details relevant for the image I mentioned in my previous message are: - D70 in bulb mode using IR remote and external shutter (very similar to your setup). I built the external shutter myself. It's primitive but it works well down to ¼ second and for flash sync. - Nikon Plan FLUOR 40x with NA 0.75 (corrected for up to 25mm FOV) - Nikon E400 scope - Koehler illumination using brightfield with an Achromatic condenser NA 0.90 - Trinocular setup (of course) - I use Nikon's RAW format (NEF) - Vignetting is cropped away using batch processing using Nikon Capture 4.3 software - Image is resized and cropped to match the output of the Coolpix 990 (see http://www.pbase.com/overney/technology_photomicrography for a comparison between Coolpix and D70) - JPEG generation is done using "normal" sharpening with Nikon Capture 4.3 software Regarding your question whether the D70 has a true 3k x 2k resolution, I must say no. It has a Bayer filter mounted on top of the CCD and hence it requires some fudging in firmware to recreate a 3k x 2k image with the correct intensities for each pixel. Such an approach is noticeably worse than a 4MPixel camera using three CCD image sensors. But I agree with you that even when using a Bayer filter, the D70's resolution is enough to sample the entire FOV when I use a 40x objective with an NA of 0.75. Cheers, Gregor |
Reply to this Message |
 Sign In
 Join
 My Latest Posts My Monitored Threads My Blog My Photo Gallery My Profile My Homepage Start New Thread Enable EMail Alerts Rate this Thread
|
©2009 Advenet LLC Privacy Policy - Terms of Use
This website includes both content owned or controlled by Advenet as well as content owned or controlled by third parties.