Photo Method for Showing Color Bias and Light Transmission

[henry link]

This method was already under discussion in this thread:

http://www.birdforum.net/showthread.php?t=130629

I would suggest reading that first for the many excellent comments. Ed (elkcub) and Ron (Surveyor) in particular have produced elaborations on the test that go beyond anything I’m equipped to do.

My original idea here was to produce a simple “picture” of color bias and light transmission that gives the viewer an experience similar to comparing paint sample chips at a paint store. It’s very difficult to judge exact color hues and light values in isolation but when two very similar colors are placed next to each other the differences become obvious.

The technique is illustrated in the sequence of photos below beginning at the left.

1) A white piece of paper (backside of a piece of Canon Photo Paper) is photographed in sunlight with a Nikon D40.

2) A binocular (old Swarovski 8x30 Porro with early “Transmax” multicoating) is positioned between the camera and the paper with the objective end pointing toward the camera and a second photo is made using the same exposure.

3) A crop is made of the circle of light that has passed through the binocular optics and that crop is superimposed on the original photo of the paper in PowerPoint and saved as a jpeg. Hopefully the central square in the PowerPoint slide looks yellower and darker than the surrounding original color on everyone’s monitor. If not, this isn’t working.

As was shown on the earlier thread, a number of binoculars can be tested together and images juxtaposed in various combinations for comparison. My hope is to continue adding test results of different binoculars to this thread, starting with a high end test to include Zeiss FL, Nikon SE, Swarovski EL and Leica Ultravid. That may take another week or so, but I’ll be happy if someone else beats me to it. Anyone is welcome to post results using this test here.

[Fireform]

I missed the earlier thread until now. Fantastic idea. I look forward to seeing the results.

[Surveyor]

Henry, some very quick, rough numbers for your sample pictures. The color shift looks similar, a little less, to the Nikon E's. The transmission appears way to high to me. This may be as simple as having to keep the camera and binocular objectives close together to keep light from outside the bino field over exposing the desired light chip segment. You may comment on your recollection of the spacing or field of view.

Ed may come up with better RGB values than I can, I do not have Photoshop and am sampling with a Icon maker program that probably is not as good.

Best
Ron

[henry link]

Ron,

I'm pretty confident about the color bias accuracy of this method, but I wonder about completely accurate light transmission. It probably will be difficult to get consistent exposure on different days and small differences in the angle of light bouncing from the paper and the axis of the binocular optics may throw things off. I think the binocular was farther from the paper today and, possibly I used a different lens on the camera, something I wouldn't have done if I had been paying enough attention. I think of this as a work in progress.

The slide below was made by combining the samples of Zeiss FL, Nikon SE and Nikon E from the earlier test photo (in the same arrangement as before) with the background color and sample of the Swarovski (upper left) from the test photo I made today. Perhaps you or Ed can determine how consistent the light levels are.

Henry

[Surveyor]

Good point Henry, I keep overlooking how bad directionality messes up level readings until I make a few, then it comes back to me in a big hurry. No way to stick a cosine diffuser on a camera that I know of.

My slow computer is in the process of imaging the drive so I just made some hand calcs. and included the RGB values so Ed, or someone, could check them. I used the RGB as D65, gamma 2.2 to convert to CIE 1931 xy and CIE L value, also shown, the approximate CCT.

I will try to plot the coordinates on the CIE chart in the morning, I have that at work.

Have a good night.
Ron

[elkcub]

Oh, I posted on the other thread a few minutes ago, so I'll just copy it here. At least for RGB values, Microsoft and Adobe software produce the same screen results. The Art Director's Toolkit can be obtained here for Mac, but Windows is also available:
http://wsidecar.apple.com/cgi-bin/nph-reg3rdpty2.pl/product=00728&cat=10&platform=osx&method=sa/downloadsoftware.cgi?id=6&ref=APPL

Very promising, Ed! I hoped somebody who actually knows what he's doing with cameras and computers would run with this.

I wonder if your software could be used to correct the overall blue bias in the test by dropping the blue channels in the image by 6? (Is channels what you call them?)

Today it's car repair. I hope to make some comments this afternoon or maybe I'll just let everyone else continue to do all the work for me. :-)

Henry

Henry/Ron/Ronh,

It's difficult to simply subtract 6 units on the blue channel because the pixels vary within each area. However, using M/S Word (of all things) I constructed a "table" with cells given RGB color values corresponding to my earlier post. From there it was an easy matter to construct a second table in which each of the B-values was dropped by 6 points. Visual differences are quite subtle. (See attached .pdf and .jpg files. The former is easier to read.)

I'm not completely sold that this manipulation is correct, because it assumes that the joint effect of the binoculars and camera is strictly additive. If you are able to take color balance readings of the white card at different times of day, as Ron suggests, it might be possible to predict the RGB values seen through the binoculars if the assumption is correct.

Like Ron, I've also found very high consistency in the RGB readings across application programs. There is a wonderful little tool called the "Art Director's Toolkit" that I would recommend for making instant color evaluation of the screen display, and also providing various conversions including CIE values, etc.

Ed

PS. Correction: the B-value in the original background should read 227 not 226. :(
... and I've added the corrected Word file.

[elkcub]

Henry,

In post #1 the background RGB is (225, 224, 228) -- slightly blue
the center of the inner box is (217, 212, 198) -- red-green

In post #4 the values are:
Background (220, 219, 223) -- slightly blue, and 220.7/225.7 = 98% of the background in Post #1.
Upper left (216, 213, 197) => 94.8% transmission (relative to 220.7)
Lower left (170, 168, 153) => 74.2%
Upper right (212, 214, 213) => 96.5%
Lower right (214, 214, 214) => 96.9%

(Assuming I made no transcription errors.) The two measures of the Swaro (or anything else from the photo) may differ because pixel color values vary within the area. I suspect this originates from vignetting, since it is typically lighter in the center and darker at the edges.

All of these single pixel measures were made directly with ADT. The only added value of Photoshop, which I used yesterday, might be to obtain averages and variances over a screen area. But it's complicated and not worth the investment just for that purpose. ADT costs $39 and has other uses as well, like making dimensional measurements on screen.

Ed

[henry link]

Thanks very much, Ed and Ron for all your efforts. Looks like some refinements are in order. In hopes of getting a more evenly lit background I'm going to try a different paper with a softer duller finish and perhaps switch to diffused northern sky light instead of direct sunlight. Then I'll keep everything parallel as best I can and try to match the crop of the background seen through the binoculars to exactly the same area on the paper. As always, the devil is in the details.

I think the old Swarovski 2 layer Transmax coating actually does have very high light transmission at yellow wavelengths (the claim was 99.5% per glass surface) but obviously it's not so good toward blue. For my eyes the the overall appearance of that square is yellowish but also darker than the broader band modern coatings on the Zeiss FL and Nikon SE. I also see the Zeiss as slightly green and the Nikon as slightly red on my computer monitor and also when I look through the binoculars. I think it's likely that the peak transmission in the SE is in red, but at a shorter wavelength than the R channel and that's why it's visible to the eye in the photos, but doesn't show up in the numbers.

Thanks again for your input.

Henry

[Surveyor]

Last night, I took a little time to take some pictures to get a better idea of the problems involved.

The test 1 set is a group of 8x20 optics, some of which I have either professional or self-made transmission data for.

The light source is a commercial, regulated lab quality halogen light with a CCT specified as 3423K. Of those binoculars tested for transmission and color, the average CCT was around 5100K.

These shots were made with the camera on auto white balance and auto exposure. Clearly useless.

Test 2 photos were made with the camera white balance set to incandescent, manual exposure of 1/10 second at f:/2.8, otherwise, the same as Test 1 series. Neither of the cameras used would allow me to completely disable the white balance.

It appears to me that I am going to have to find a camera that I can completely disable the white balance. It seems clear to me that the camera software is adjusting whatever it deems as correct exposure to 6500K (D65) so even minor intensity changes alter the output color.

I have thought of two other procedural changes I am going to try this weekend if I can find a camera that I can disable the white balance on. Of all the cameras I have, I am not sure I have one that can go completely manual and disable the white balance. One is instead of a target sheet, I am going to change to a 6x6” ground glass target for a more diffuse light source. Next, I want to try using a very collimated monochromatic light source, a 532nm and 640 nm laser.

Ron

[elkcub]

Good point Henry, I keep overlooking how bad directionality messes up level readings until I make a few, then it comes back to me in a big hurry. No way to stick a cosine diffuser on a camera that I know of.

My slow computer is in the process of imaging the drive so I just made some hand calcs. and included the RGB values so Ed, or someone, could check them. I used the RGB as D65, gamma 2.2 to convert to CIE 1931 xy and CIE L value, also shown, the approximate CCT.

I will try to plot the coordinates on the CIE chart in the morning, I have that at work.

Have a good night.
Ron

Ron/Henry,

Looks like we get about the same RGB readings, -- see my post #5. What is your motivation for converting to CIE values?

After taking a spot reading from a screen area (or an area average), it seems to me that then creating a uniform color swatch with these values is about the best one can do to model, document or compare the colors side by side. This will always be an approximation, however, because a screen image is decoupled, and, as you suggested in earlier posts, will be affected by the screen properties.

I don't think one can read too much in to the RGB values, per se, because the individual's unique retinal sensitivity function conditions the perceived color of the swatch when seen on a screen. Still, I continue to think the method may be worthwhile for making general comparisons.

Ed

[Surveyor]

Hi Ed;

CIE, like a lot of bins, is just a personal preference for me. I find it easier to visualize 2d coordinates than 3d. I also need CIE to do the color temperature conversion, and again, find it easy to visualize a one-dimension value over a 2d. A lot of this is left over from my old days as a radio-TV station engineer and some other factors, which are:

CIE is the standard for “absolute” color chromaticity values, or at least the industry standard. They stand alone, any xy pair should be the same as another same pairing. In contrast, RGB is a device specific standard so manufacturers can bring there devices to show the same colors. The CIE tristmulus XYZ are constant for a given integration of a transmission curves component powers and others systems are modified to display that color. If the device dependent displays are set to a different system or color space than the originating system, the viewed color will be a little off and, even though the RGB values are the same on both monitors, if the color temp or, more specifically the gamma, is different, the color viewed will be different. The monitor I am using at the moment, an old Sony Trinitron, only has two presets, 9300K and 5000K, gamma 1.8, so the chances are that I will see different colors than you if your monitor is set at 6500K gamma 2.2 even though the picked RGB are identical. This is easily demonstrated if you are using a monitor with a front panel menu system. With a color screen displayed (RGB’s set by the software) change the color temp setting and watch the colors turn dimmer, yellow with lower settings and whiter/bluer with higher settings. The RGB values will not change though. You can then see how people viewing a posting with monitors at various settings, a lot just set to what the individual likes to see, will be seeing different colors than you. Example, if my monitor were set to 5000K, I might call a swatch yellowish and someone with the monitor set to 9300K or 11000K might see the same swatch as blue-green.
All of my test equipment, spectroscopes, calibrated light sources, monochromators, luminance meters, spectrometers, etc. are calibrated in CIE values. The only time I deal with RGB is with device settings to match components like printers, scanners, monitors, cameras etc. Then I usually only match the color temp settings. Example a camera may be set for incandescent light in (3500K) and sRGB D65 output so I would need to make sure my photo processing software (or other graphics software) was using these settings to get the proper color prints. If the devices could faithfully process CIE coordinates, that would never be an issue, but since you have to tell a device what a color is, you need some way to communicate that information.
All of the test data I have available, including curves from binos already tested are in CIE transmission and color format. It is just easier for me to compare the new data with the old when the formats match.All that said, I have no problem using RGB as long as the reference color space is known. That is why I have been using the defacto windows standard of D65 gamma 2.2 since a better definition was not supplied. But you still have a problem with CIE plots, the coordinates are correct but, as someone called it, the color schmoo may be off when displayed or printed.

Forgot, the eye receptor sensitivity curves found are based on CIE.

Anyway, like I stated, this is personal preference. I am not advocating using it as a forum standard or anything like that. Color is such an abstract for most of us, it seems to be more of a matter of what you get used to.

Clear as mud, right. :smoke::eek!:

Have a good night.

Ron

[elkcub]

Hey Ron,

Sorry to take so long responding. Your explanation hit the spot! :t:

Ed

[Surveyor]

This is my latest effort for the photo method. I tried a couple of various things, the photos shown in Test 5 are the most consistent to visual observations so far but still do not conform to previous tests by myself or a professional measuring lab doing NIST traceable work. An example for the Ultravid is attached and my tests were about 100K CCT higher, a good correlation since I used a different reference illuminant, D65, and the transmission curves were very similar. I have attached my color measurements for the Ultravid, I do not have the professionals results, only that shown in the transmission report, so I do not know how tightly the CRI’s matched.

Ed (Elkcub), I thought these two files would be of particular interest to you. Seem right up your alley.

This series of photos were taken with a lens between the light source and objective in an effort to better control the directional properties, but well short of trying to get a small collimated beam as you would for more serious testing. I also used a cosine diffuser between the exit pupil and camera lens, the reason for reversing the direction from Henry’s proposed positioning. The Test 4 results used the same setup but normal exposure, which resulted in everything being 255,255,255 so Test 5 was deliberately underexposed.

While these results “look” acceptable they will not match any ordering of the more formal test results and I made a conscious decision to search for the RGB value further from the Reference position although moving the “pick” point could locate almost any desired color between the two. Anyway, this appears not to be the approach needed for this to work. Further tweeks are necessary.

Back to other projects for a little while.

Best to all.
Ron

Forgot to mention, the manual white balance was set to the Reference frame for Test 4&5.

[elkcub]

I'm satisfied that RBG values remain unchanged (numerically) when I modify my computer's screen profile. Although the PowerBook G4 that I use can be changed from its default "Color LCD" setting to "CIE RGB" or "sRGB" (and many other profiles), the visual impression of subtle color differences in Henry's pictures remains largely unchanged (to me). So, for whatever reasons, I think his technique has the potential to capture valid color properties that can be shared over the Internet, at least within the range (or gamut) of the typical display. Personally, I think the effort should be to create a uniform color swatch in each case, based on central RGB pixel values, or an area average. The patch is readily made with Word or similar software. Yes, I realize this is not quite kosher colorimetry, but formal specification of the colors would be rather expensive.

Ron, you might be interested in comparing this professional evaluation of the FMC Swift 804ED done a year or so ago with your 8x20 Ultravid's. This is Renze de Vries c. 1995 specimen sent to Germany at the time. The second file is for an older 804, possibly an MC HR/5 from c. 1990.

Ed

[Surveyor]

Ed;

Thanks for the files, I will add them to my collection. On cursory examination, it appears they used a 16 mm collimated beam, illuminant E with a 3300K light source and 5 nm steps. That should be enough to compare formats. Always good to see this kind of data. Any idea what they charged Renze for this service?

My LCD screens use the same ICC and ICM profiles you mention, but I really have no idea what parameters they are setting the monitor to, there is probably a listing somewhere. The color temps and gamma's I was comparing were on CRT's that had front panel controls for those settings.

On the topic at hand, I believe you are correct about having to get to a "uniform" swatch before anything else. I found an online calculator and just made a couple of trial runs. I have not looked closely at the output yet, hope you have a chance to. I could not find a price or way to download the software but will check further on that.

The bottom line of the attached screen shots gives the RGB Average, Median, Min and Max for the swatch. I used about the middle quarter of Henry's attachment.

Note the Hue plot of the hsv section. That may end up being a better visual indicator than the RGB value.

Best,
Ron

[elkcub]

Ron,

My impression was that a 3mm beam was used. But, I speak no German, so it's all Greek to me.

What and where is the on-line calculator?

The Mac has quite an extensive ColorSync Utility, which provides all the tristimulus values and rotation matrices. I agree, that hue, saturation and intensity value (HSV) is a good reference model, particularly for human color matching. Does the on-line tool allow for this? To be useful, everyone needs such a tool.

Ed

[Surveyor]

Ron,

My impression was that a 3mm beam was used. But, I speak no German, so it's all Greek to me.

What and where is the on-line calculator?

The Mac has quite an extensive ColorSync Utility, which provides all the tristimulus values and rotation matrices. I agree, that hue, saturation and intensity value (HSV) is a good reference, particularly for human color matching. Does the on-line tool allow for this? To be useful, everyone needs such a tool.

Ed

Ed, I am going to have to go to the Google translator at home to figure it out the information needed.

Sorry, forgot the link. http://mkweb.bcgsc.ca/color_summarizer/?home
Just go to the analyze page and browse your swatch. I only played with it for a few minutes and did not try to download the results yet.

Ron

PS. I got in touch with the link owner. This is just available online (a script file) and not available as a stand alone app.

[Surveyor]

A few more tests have led to more promising results.

Test 5, the swatchs were made from a 100x100 pixel sample from the center of the image (1024x768 display mode) or about the center 1.5% of the image. Using the link above to average the area, the saturation value was low (the S part of HSV and was <5) as a result of the underexposure. These did not fit well with other data. The white balance for this test was set to the reference frame.

I went back to the Test 2 images and cropped out 200x200 pixels of the 1024x768 display (5%). These images had an average S value around 15 but the reference images were a little overexposed. The white balance for the series was set to ‘incandescent’. These ordered well compared to other data, see table.

I have attached the sampling data from the Ultravid, one sampled as described, the other sampled from the full frame to a 200x200 pixel area (0.8%).

The results of this single test indicate that the test may well be valid, but the parameters still need some experimentation for the best repeatability.

It appears to me at present that we need to sample at least 5% of the central area and keep the S value as high as practical without overexposing. I am not at all sure yet which white balance setting to use.

Ron

[elkcub]

Ron,

Since you have the 8x20 Swaro, Ultravid, and LX L, — which do you like best? In particular, I'm interested in the Ultravid vs. LX L.

I've tried to take photos of various bins on a light table with 5000K bulbs, but so far nothing comes out as well as Henry's. Basically, I just get gray, similar to yours. ;)

Ed

[Surveyor]

Ron,
Since you have the 8x20 Swaro, Ultravid, and LX L, — which do you like best? In particular, I'm interested in the Ultravid vs. LX L.
Ed

Ed,
My needs differ from most, I need true pocketability. From that perspective the Nikon LXL falls towards the bottom I the list of what I carry most of the time. As far as pocket bins go, the Ultravids and Zeiss 6x20 are my most carried with the Swaro and Trinovids next (I prefer the ergonomics of the Trinovids but the Swaros are a little better bino, I just don't like the pinky focus on it or the LXL) then the LXL and the Victory least of all.

If size is not the primary concern, the LXL's are optically as good as the Leica's, just a shade more red bias.

Ron

[henry link]

Ron and Ed have done some amazing work with the raw images from my first efforts at this test. Without their analyses I wouldn’t have understood the need for very careful control over the angle of light, evenness of illumination and the exact position of the crops. I have no ability to generate the numbers and charts they can do, so I am concentrating on trying to achieve better reliability and consistency in the original test photos. I continue to hope this test can provide a simple and direct way to make a visual comparison of binocular color bias and light transmission that the eye will instantly recognize, just like comparing paint swatches. The color part seems to be pretty easy, but light transmission accuracy and consistency are hard to achieve. It’s a continuing process. I’ve thought of a few refinements since I made the images in the photos below, but I’m going ahead and posting them as examples of the best efforts to date.

I made a trip to a dealer who is very tolerant of my geeky testing and stocks some models of the (so called) alpha brand binoculars. They allowed me to set up a test in the parking lot. It had to be in direct sunlight, but I’ve tried my best to produce evenly illuminated crops of the same area of the paper in each binocular. In every binocular’s color swatch I had to crop out a bright upper left corner, so I suspect the main culprit in unevenness was a spot of glare on the paper itself, which was imperceptible to the eye.

The left image shows a comparison of four binoculars: Leica 8x42 Ultravid (upper left), Zeiss 7x42 FL (upper right), Nikon 8x42 LX-L (lower right) and Swarovski 8.5x42 EL.
The right image shows a comparison of a Zeiss 7x42 FL (top) and a Zeiss 8x32 FL. I thought that one would be interesting since the optical design and coatings of those two are nearly identical except that the 7x42 uses an Abbe-Koenig prism and the 8x32 a Schmidt-Pechan.

Naturally I have my own opinions about what I see in the images, but I would be very interested to hear what others see in them first (and of course inscrutable charts and numbers from Ron and Ed will be most welcome also).

[mooreorless]

Hi Henry, I have been following most of this and I hope you don't mind me saying what I see with my monitor. The first images I see both on left Leica and Swaro are very close and looks slightly grey bias, the right upper Zeiss Fl is white the right lower is red bias"Nikon". Second image top white and lower slight grey. I tried to do this without remembering where the images were for each binocular.
Thanks for taking the time to do this.:t:

Regards,Steve

[ronh]

I once had occasion to radiation-damage antireflective layers of magnesium fluoride on glass (related to protective covers for solar cells on satellites), and then measure the loss in transmission vs wavelength in the damaged spot. It was easy to see a spot having only 1% transmission loss in the visible. From that, I would estimate the difference between the AK and SP FLs to be about 1%. I think it's said by Zeiss to be 2%.

This is a very sensitive test, so sensitive that it could lure somebody into becoming overly fixated on trivial transmission differences. Hey, that's a compliment!
Ron

[Surveyor]

Henry,

I first tried to get a feeling of color shift just by visual appearance as you requested. My estimates for Slide 2 were that the shift from the background of the two left swatches either were down in transmission or had a shift towards the red end of the spectrum and since I knew the characteristics of the two binos, I estimated a shift towards red but could not tell how much. Looking at the top right swatch, I estimated a shift from the background in the blue-green direction; I was able to cheat a little here since I know all these models are within a few percent of each other in transmission. The bottom right swatch I estimated as having less transmission and a greater shift in the red direction than the two left swatches.

Then I went back through them, measuring the RGB values as previously described with the following results:


Background average of 8 samples=227.5,229.375,231.5
Leica 8x42 Ultravid value=215,215,215 and shift=241,239,237
Swarovski 8.5x42 EL value=215,215,215 and shift=241,239,237
Zeiss 7x42 FL value=215,219,218 and shift=241,243,240
Nikon 8x42 LXL value=212,213,208 and shift=238,237,229

This leads to the following estimates of transmission and color shift:

Leica 8x42 Ultravid = transmission 95%, shift 30 degree, S=2 (orange)
Swarovski 8.5x42 EL = transmission 95%, shift 30 degree, S=2 (orange)
Zeiss 7x42 FL value= transmission 96%, shift 100 degree, S=1(light green)
Nikon 8x42 LXL value= transmission 94%, shift 53 degree, S=4(orange-yellow)

Note that the saturation number may end up being analogous to value of intensity along the dominant wavelength in CIE colormetry i.e. the Zeiss only shift 1% in the light green direction, while the Nikons shift 4% in the orange-yellow direction. This is only a thought and needs further investigation.

Henry, you might clarify one point. My colormetry work starts with a known value and mathematically arrives at the native color point of the bino. I am assuming that you intend to estimate the shift from whatever background is present, instead of a fixed value. I went towards the shift perspective since it looks as if it matches most perceptions better.

Now on to Slide 3. Visual inspection leads me to think there is a color shift of either orange or yellow from the background with a small transmission loss in the bottom swatch compared to the top (more about this later).

The measured parameters were:
Background average = 228,228,232.5
Zeiss 7x42 FL value = 219,221,218 and shift =245,247,239
Zeiss 8x32 FL value = 213,215,212 and shift = 238,240,233

Resulting in:

Zeiss 7x42 FL transmission (Luminance) 97.1% shift 75 deg. S=3% (yellow-green)
Zeiss 8x32 FL transmission (Luminance) 94.9% shift 77 deg. S=3% (yellow-green)

At this point, something I have been thinking about seems to become more obvious. This method is not measuring transmission, but rather the luminance of the swatch and would, therefore, be dependent on the aperture and transmission, maybe not an indication of true transmission.

Henry, you seem to be controlling direction and intensity very well for an outside, illuminated flat sheet. How are you doing it??? This appears that it is going to be a valid test for comparisons done under the same conditions at the same time. It may even carry to independent status under more stringent conditions, but that may negate the simplicity sought.

I have never done any formal testing of the Zeiss FL series, so I am not familiar with what bias I would perceive in the view.

I, also, tried to stay within the confines of RGB, at least to the limit of my understanding of that system since I almost never use it. Ed will need to verify my RGB perceptions (I hope I did not mess them up to bad).

Best
Ron

[henry link]

Ron,

Thanks for all your work, again.

The first thing I notice is that the background color and FL crop from the two slides don't generate exactly the same RGB numbers. I don't understand that since they are the very same images transferred from iPhoto to PowerPoint, though I may have resized and reshaped them slightly differently once they were transferred. Should we just expect some small differences in each transfer or should I avoid any resizing or shaping?

As far as controlling intensity, I had to resort to very small crops (about 1" square) inside a 4" square I had penciled onto the background paper. I found I could see small differences in intensity within a crop by transferring a duplicate of it to PowerPoint, then juxtaposing the original and duplicate with different rotations and overlaps. I moved and adjusted the crops until I finally found a crop of pretty even intensity. Then I tried to match that with the same area on the background. Quite a pain in the rear.

Today I tried something different, which is illustrated in the slide below. I used the controls in iPhoto to brighten the four "alpha" binocular samples so that they all approximately match the background brightness which leaves only the pure color bias visible, like the smile of the Cheshire Cat. The iPhoto control of "exposure" assigns numbers to the increases. I don't know what the numbers actually mean, but this is what was required to bring each binocular up to approximately the brightness of the background: Zeiss FL-.09, Leica Ultravid-.13, Swarovki EL-.15, Nikon LX-L-.17.

Also, thanks Ron and Steve for the visual impressions. I think the Zeiss, Nikon difference is a pretty good test for red/green color blindness. And thanks, Ron #2, for the compliment only a true optogeek could appreciate.

Henry

[Surveyor]

Henry, the problem with exact matches of the RGB numbers is that I am averaging an area to get the value and even trying to pick the same area (in the above case I used 50x50 pixels) you will get a difference of a couple of counts. I do not have the time at the moment to redo the last Slide you posted, but I did crop an area out of the Nikon, Zeiss and Background of the last slide and process with the online software I listed in Post #17 (since this is an online script, I think it will work with Macs). I cropped a large section of each patch to a separate jpg and loaded that into the program.

Look at the top three rows, the h, s and v rows and, also, notice the color wheel plot on the left of the page. Note how when sampling the patch the RGB values range in color and amplitude of various widths. The upper left box with a saturation value directly below shows the average hue. The total deviation from white is the fourth row, HSV value. The RGB values are below that, but it is hard for me to visualize them.

At the moment, I do not see this as much of a factor since we are looking for a shift from the background to the inter-optic patch.

When you did that last manipulation, it really brought out the red and green, at least on my monitor.

[elkcub]

With reference to Post #21, as Ron mentioned the various areas are not quite uniform. The following RGB readings, therefore, are estimates of an average value.

Four section panel:

Background 227, 228, 232 (blue bias)

Upper Left 216, 216, 216 — 8x42 Ultravid
Lower Left 215, 215, 215 — 8.5x42 EL

Upper Right 218, 220, 217 — 7x42 Zeiss (slight green)
Lower Right 215, 212, 207 — 8x42 LX-L (slight red-green)

The two section Zeiss FL panel:

Background 227, 228, 232 (same blue bias)

Upper 218, 220, 217 green bias, (2% greater transmission?) — 7x42 FL
Lower 213, 215, 212 green bias — 8x32 FL

Comments about color tint are based on the RGB weights, but also correspond to my subjective impressions of the various swatches in the panels.

The way I tend to think of this is that the background models daylight, perhaps as seen by the eye, and the various swatches tell us how this is modified by the instrument. This is the work of the optical transfer function for each instrument. So, for example, the Ultravid and EL filter out some blue to produce an equally weighted RGB, whereas the FLs, and LXL change the bias from blue to green or red, respectively. With the appropriate rotation matrix, these values could be represented in HSV format that Ron prefers, but the information content remains the same.

Ed

[elkcub]

Then I went back through them, measuring the RGB values as previously described with the following results:


Background average of 8 samples=227.5,229.375,231.5
Leica 8x42 Ultravid value=215,215,215 and shift=241,239,237
Swarovski 8.5x42 EL value=215,215,215 and shift=241,239,237
Zeiss 7x42 FL value=215,219,218 and shift=241,243,240
Nikon 8x42 LXL value=212,213,208 and shift=238,237,229

This leads to the following estimates of transmission and color shift:

Leica 8x42 Ultravid = transmission 95%, shift 30 degree, S=2 (orange)
Swarovski 8.5x42 EL = transmission 95%, shift 30 degree, S=2 (orange)
Zeiss 7x42 FL value= transmission 96%, shift 100 degree, S=1(light green)
Nikon 8x42 LXL value= transmission 94%, shift 53 degree, S=4(orange-yellow)


Ron,

I'm having a senior moment. Don't follow how you arrive at shift values in RGB. I understand the H-shift in degrees when comparing two hues.

Ed

[Surveyor]

Hi Ed;

Hopefully I haven't derailed myself with the RGB. What I did was divide the patch R, G and B by the Background R, G and B and multiply by 255 so I could directly compare with the white value of 255,255,255. Example, the Leica and Swaro came out 241,239,237 compared to 255,255,255 white for both luminance and color shift.

Best,
Ron

[henry link]

Thought I would post one more image which shows some interesting changes in the color bias of the Zeiss FLs since they were first introduced. The photo below shows a 8x56FL from about 2 years ago (upper left), the 7x42 FL I tested at the dealer last week (upper right), the Leica 8x42 Ultravid from the same test (lower right) and an early production 8x42 FL from 2004.

To my eye it appears that the early FL had a more neutral appearing color bias which more closely resembles the Ultravid. I noticed the slighty greener bias in the 8x56 FL when I first bought it. It's an early example of a LotuTec coated FL and my suspicion fell on that as a possible cause. It's also possible that an unrelated change in the coatings was made to extend high transmission further into the blue/green for a better match with peak eyesight sensitivity in low light. Whatever the reason, it appears that recent FLs have a greener bias than early ones. On my monitor the new 7x42 appears obviously greener than the 8x56 from two years ago.

[elkcub]

Hi Ed;

Hopefully I haven't derailed myself with the RGB. What I did was divide the patch R, G and B by the Background R, G and B and multiply by 255 so I could directly compare with the white value of 255,255,255. Example, the Leica and Swaro came out 241,239,237 compared to 255,255,255 white for both luminance and color shift.

Best,
Ron

Hi Ron,

I think you're right on track. You expressed the observed R, G, and B values from the binocular-plus-camera as a proportion of the R, G, and B values obtained from the camera alone. When these proportions are taken of pure white (255, 255, 255) it models the binoculars' (pure) color shift. Of course, this assumes constant proportionality, which makes sense.

To get an idea of what this tint (241, 239, 237) looks like one can use the RGB calculator at: http://webdeveloper.earthweb.com/repository/javascripts/2001/06/48571/hex.html. In HSV terms this comes out to (21, 4, 241). (Unfortunately, the calculator only produces .tiff files, which BF doesn't accept.)

But, our calculations don't quite agree. Using your RBG numbers I would calculate the binoculars' transmission as 215/229.45 = 93.7%, where 229.45 is Sum(RGB)/3. Also, the calculator's HSV is 21 deg. not 30. What am I missing?

The basis for my RGB preference is that it is essentially the Young-Hemholtz theory. A good summary of this and the many normalizations and transformations later developed by psychologists and CIE can be found in "Optics" by Freeman and Hull. http://books.google.com/books?hl=en&id=lXNFnybj9wwC&dq=optics+freeman+and+hull&printsec=frontcover&source=web&ots=P6ica2IRcT&sig=HwiLJ1xSgankWiMgADnGY-MV1CI&sa=X&oi=book_result&resnum=1&ct=result#PPA368,M1 Unfortunately, the whole section on colorimetry isn't available to read on the Internet, but at least some of it is. (I bought the book some time ago.)

Ed

[elkcub]

Thought I would post one more image which shows some interesting changes in the color bias of the Zeiss FLs since they were first introduced. The photo below shows a 8x56FL from about 2 years ago (upper left), the 7x42 FL I tested at the dealer last week (upper right), the Leica 8x42 Ultravid from the same test (lower right) and an early production 8x42 FL from 2004.

To my eye it appears that the early FL had a more neutral appearing color bias which more closely resembles the Ultravid. I noticed the slighty greener bias in the 8x56 FL when I first bought it. It's an early example of a LotuTec coated FL and my suspicion fell on that as a possible cause. It's also possible that an unrelated change in the coatings was made to extend high transmission further into the blue/green for a better match with peak eyesight sensitivity in low light. Whatever the reason, it appears that recent FLs have a greener bias than early ones. On my monitor the new 7x42 appears obviously greener than the 8x56 from two years ago.

Henry,

Here's what I get:

Background 229, 228, 232 C= Sum(RGB) = 689

UL 219, 221, 220 Trans= 660/689 = 95.79% 8x56FL
UR 218, 220, 216 Trans= 654/689 = 94.92% 7x42FL

LL 218, 218, 218 Trans= 654/689 = 94.92% 8x42 Ultravid
LR 216, 216, 216 Trans= 648/689 = 94.05% 8x42 FL

Again, there is some variation in each panel. Transmission values are too close to tell apart by eye, IMO. Ron can elaborate on color shift, but I would say the early FL and Ultravid are more "neutral" in the sense that they produce white from daylight. And, yes, the newer ones have a greener bias — good eye!

Ed

[henry link]

Thanks for the measurements, Ed.

I should have mentioned that in order to compare the results of the earlier test of the 8x56 and 8x42FLs to the test of the 7x42 FL and 8x42 Ultravid done at a different time and place I had to resort to equalizing the background "exposure" values for the two tests in iPhoto and then applied the same equalization to the crops. Changes in the angle of the sun cause the test brightness values to wander. I may have cooked up a fix for that. I plan to glue an object to the edge of the paper and trace its shadow with pencil. Then all subsequent tests will be set up for the sun's shadow to match the tracing. Then, of course I still have to get the binocular optical axis perpendicular to the paper, something I've just been eyeballing so far.

The biggest surprises for me in the light transmission numbers that you and Ron generate are how very high they are and how closely clustered. In this last test, for instance, I have no trouble seeing the difference in brightness between the FL and the Nikon LX-L by just looking through the binoculars, I would have guessed it's between 5-10%, closer to 10%. The tiny differences your numbers indicate are not something I expect to be able to see at all. From published tests I would have guessed the FL would have transmission around 94-95% (440-450nm), the Ultravid and EL around 89-91% and the LX-L maybe 86-87%.

I was also very surprised to see the closeness of the Ultravid and EL in both color bias and light transmission. They look like they could be two samples of the same binocular.

Henry

[elkcub]

Thanks for the measurements, Ed.

I should have mentioned that in order to compare the results of the earlier test of the 8x56 and 8x42FLs to the test of the 7x42 FL and 8x42 Ultravid done at a different time and place I had to resort to equalizing the background "exposure" values for the two tests in iPhoto and then applied the same equalization to the crops. Changes in the angle of the sun cause the test brightness values to wander. I may have cooked up a fix for that. I plan to glue an object to the edge of the paper and trace its shadow with pencil. Then all subsequent tests will be set up for the sun's shadow to match the tracing. Then, of course I still have to get the binocular optical axis perpendicular to the paper, something I've just been eyeballing so far.

The biggest surprises for me in the light transmission numbers that you and Ron generate are how very high they are and how closely clustered. In this last test, for instance, I have no trouble seeing the difference in brightness between the FL and the Nikon LX-L by just looking through the binoculars, I would have guessed it's between 5-10%, closer to 10%. The tiny differences your numbers indicate are not something I expect to be able to see at all. From published tests I would have guessed the FL would have transmission around 94-95% (440-450nm), the Ultravid and EL around 89-91% and the LX-L maybe 86-87%.

I was also very surprised to see the closeness of the Ultravid and EL in both color bias and light transmission. They look like they could be two samples of the same binocular.

Henry

Hi Henry,

A basic question, I think, is the validity of what we're doing. By that I mean, how do our inferences from simple pixel values correspond with refined spectral transmittance measurements of luminous flux? At best, I would guess they are a gross approximation, which hopefully retain ordinal relationships, i.e., A > B > C, etc. For several reasons I can think of its probably too much to expect the scaling to be linear. Ron, no doubt, might comment more about this.

Ed

[Surveyor]

Ed,

Sorry to be so long getting back to you but I am having computer problems at work and have to deal with the stuff that pays the bills first. Concerning post #31, I am glad I did not stray too badly. About the difference in the HSV, I think this probably concerns settings of the converters. I have attached a jpg of the calculator I use and for HSV I use the format 0-360 degree for H and 0-100% for S and V. The converter I use will allow me to select from 6 formats for HSV. From the results you show, I think your calculator may default to 0-255 counts for all three values (21/255=30/360, 4/255=2% and 241/255=95%). The difference we have in T% or luminance is that you are using the straight RGB values and I used the CIE luminance value, or L in the Lab definition. The numbers will track in the relative sense, but does point out the need for a good protocol definition.

Henry, ED;

Like Henry, I am surprised by the high values for the assumed T%. At this point, I am thinking what we are seeing is an average of three lens systems, the bino, the camera and the CCD photosites. We are probably also getting an electronic gain figure from the camera program that charges the CCD for a given exposure setting. If this is indeed the case, I think the T% values are more along the lines of a random distribution and are going to be undependable for the use intended, but this is strictly conjecture at this point.

Based on the procedures I have to go through when doing transmission measurements, I am going to assume that a simple transmission estimate is fairly unlikely. Controlling directionality and intensity setup take far longer than actually doing the measurements. For instance, a tightly collimated beam with a maximum dimension of ½ the aperture and preferably less controls directionality. Anyone who has tried to collimate a visible light beam down to close to laser dimensions will understand the problems. Controlling the intensity is a set of problems all their own.

I have attached a jpg of three of the binos referenced (although 20mm versions) that show the levels we should be getting close to but I have never had the chance to test any of the FL range. The values Henry referred to would be typical peak values. The average values for the range of 380-780 nm would be:

Ultravid=76.05%
Swaro=69.77%
LXL=79.78%

For the photopic range of 500-600 nm (I assume camera sensitivities peak in this range, but do not know that for fact):

Ultravid=88.7%
Swaro=87.8%
LXL=85.6%

The color difference is another story. These appear to be tracking very well to me. As Henry noted the difference between the Swaro and Ultravid is very close to the same. Look at the transmission curves and this is confirmed, the shapes are remarkably similar. The colors from the color test were very close also, the color temperature only differed by 11K. You can see the red bias in the LXL plot, very consistent with Henry’s photos.

Still, a method that deserves more study and experimentation.

Best to all.
Ron

[elkcub]

Yup, that makes sense. The calculators can be scaled differently within the same measurement framework.

But, I suspect that the computed GRB transmission values were also distorted by Henry's iPhoto correction of the "exposure". This may very well have had a disproportional impact on relative magnitudes because the upper numerical boundary is 255. Of course, I can't prove it.

Henry: In #33 did you really mean 440-450 nm?

Ed

[rjwims]

Why did this thread die?

Seemed to be moving along nicely...:-C

Thanks