Magick Numbers

Hi all,

A picture is worth a thousand words (and a million numbers) but sometimes diving deep into technology can help you use technology at greater depth, at least that was the idea. So I have been doing some spectral measurements of magick filters combined by computer simulation of transmission of light in pure seawater to get a better feeling for what is going on. If you dive in real tropical blue water, temperate green water, or brown mudpools you will obviously get somewhat different results but this still gives a good sense of the magniture of water and filter effects.

Scientists have measured and tabulated the transmission of the entire visible light spectrum in water

(R. M. Pope and E. S. Fry, "Absorption spectrum (380-­700nm) of pure water. II. Integrating cavity measurements," Appl. Opt., 36, 8710--8723, 1997). I've taken that data and calculated the effects on the spectrum of light at 1, 2, 4, 8, 16, and 32m depth. The results are in the next image.

Simulated water light transmission spectra at different depths

The next image shows the wavelength sensitivity of a typical Bayer Filter used to form the red, green and blue pixels in CCD cameras.

(taken from: http://www.molecularexpressions.com/primer...agesensors.html which is worth a read if you like technology)

It is clear that below 580nm water transmits light very well, but this gets rapidly worse for longer "red" wavelengths. If in the first figure you draw a horizontal line through the 0.5 transmission point it will intersect the 2m line at about 650nm. If we take this as our "average red" then it means that for each 2m of water you lose one stop of light. This is important because it gives an idea of how deep we might be able to push filter photography. Basically to shoot N meters deep you need to overcome N/2 stops of light loss. Some of this can come from your camera white-balance. If you can get away with a 3-stop (or 8-fold) boost of red using white-balance then you can do without filters up to 6m depth. (A 3-stop boost corresponds to "sacrificing 3 of the 12 bits dynamic range available per color). To go beyond that you need to use aperture, shutter time, and ISO setting to boost exposure. To make that work you need to avoid overexposure of the blue and green colours by using filters.

To get the ideal filter at a given depth you want the transmission of the filter multiplied by the transmission of the water at that depth to give a "neutral effect" on the spectrum. In other words, you want to change the curve in the first figure into a straight line. The next figure shows idealized filters that do just that and have (unrealistic) perfect transmission at long wavelength. Because there really isn't any practical amount of red left at 32m, I'm only showing optimal filters for 1 to 16m depth. In addition, I've choosen to set transmission at 650nm to 1.0. To get ideal behavior for longer wavelengths would require further attenuation of the blue and green which is ok for shallower depth but is not realistic for 16m. Again you can see that you need a 1-stop (two-fold) attenuation of blue and green for every 2m of depth. For 16m, the blue and green have to be attenuated by about 250-fold, or 8 stops! I don't think that is impossible but clearly autofocus and adequate exposure are going to be challenging.

Idealized filters for different depths to give a "neutral spectrum" up to 650nm

So what does the magick filter transmission spectrum look like. To find out I measured a wavelength scan as shown in the next figure.

Magick filter transmission spectrum

The filter does a very good job at letting through the red light with a steep climb in transmission just around 590 where water absorption causes a deep drop in transmission. Compared to the idealized filters the Magick filter clearly lets through a lot more green than blue. This is not optimal with respect to the pure "laboratory" seawater. Another thing to note is that the maximum attenuation is just over 2 stops which according to the simulations compensates for about 4m of water. Combined with a 4 to 6m water-compensation from the camera's white balance that would put the sweet spot for filter photography in the 6 to 10m range. To go deeper you can try to use two filters or, better, one magick filter plus a filter that brings down the green part of the spectrum.

HOWEVER, before rushing out to buy more filters please do remember these are simulations. Algae and particulates in real seawater also absorp light and maybe they do so predominantly in the green part of the spectrum making the magick filter just perfect. I'm planning to do some real-world experiments in May in Caribbean waters. That is, if I don't get too excited by just taking pictures of fish with my new toys. After all, in the end a picture is still worth more than a thousand words!

Bart

Great stuff, Bart.

Thank you very much for posting this. I certainly find it very interesting, to see the numbers behind the Magic.

Here are some quotes from the Magic Website (written last year) that show that experience using filters ties in with your findings:

Traditional underwater filters are also designed to exactly counteract the filtering effect of seawater. The problem is that the filtering characteristics of water change with depth and with what is in the water - so such filters only work well at a specific depth in very specific conditions.

 

As photographers Peter and I knew that we didn't want to stick a filter on our camera that wouldn't work above 5 m (15ft) because that would guarentee the appearance of mating whalesharks right at the surface!  Particularly with moving sujects you can never be sure exactly what depth you will encounter them.

 

The Magic Filter was specifically designed to work over a wide depth range (0-15m/0-50ft) and will produce useful images down to more than 20m.

I'd also add that the decision not to entirely balance the green part of the spectrum was a concious one. It is relatively easy to make a filter that balances the spectrum, but the problem is that if you do this the filter attenuates much more light and also it will not work close to the surface. Also in field tests we found that the camera was able to deal much more effectively with too much green, rather than too much blue. Maybe this is because digital cameras are programmed to deal with shooting in fluorescent lighting that has a strong greeny tinge?

Anyway. although the filter was developed to produce the goods while diving, I am glad that it also makes sense in the laboratory.

Alex

p.s. I will add some links to your calculations on my own and the Magic wesbites.

And to brighten this thread up, here are some pix:

Alex

Isn't there a WORD LIMIT on this site. If a picture is worth a thousand words than that last one was at least 9,000 words long, EXCESSIVE !

Interesting comments about the response to green light. The Bayer filters also dedicate twice as many pixels to green than to red or blue. Never understood that really but perhaps there is something special with green. It is the colour that human eyes are most sensitive to, so loss of green may have more of a visual impact than other colours.

Another thing to keep in mind is that although light comes in a continuous spectrum, our eyes and cameras capture the colour as just three "values" red, green and blue. Although correcting the entire spectrum is ideal, in practise it may be good enough to get the relative values between the 3 colour pixels right. E.g. a lot of red in the 600-650 range may compensate somewhat for a lack of red in the 650-700 range. That is exactly what the field tests should show as I'll be using the actual camera to make the measurements.

For my field test report I'll promise that I will also post my best "newbie" shots, with and without flash, which will no doubt reveal that most of the magick still comes from the skill of the magician pushing the shutter.

Bart

I think it's the model in that last photo that really contributes to the "wow" factor :-) Except for the two differently colored strobes - that's just WRONG!

Cheers

James

I think it's the model in that last photo that really contributes to the "wow" factor :-)  Except for the two differently colored strobes - that's just WRONG!

<{POST_SNAPBACK}>

Maybe the filter has a different effect at the edge of the frame compared to the middle?

I wonder if the problem would be worse on a full frame camera?

Alex

I thought so! Using the magic filter at 15+m in the Red Sea last year I ignored Alex's advice and photoshopped the images to drop the green channel, which is what you would expect from this analysis. The shallow image is basically unmodified, but the deeper image was modified in "curves" to make the sand whitish, which was done by reducing the green channel.

Tim

B)

Ok, ok, I'll be sending an order off for a set for my new Fuji E900 on my next trip to Manado in a month.

But I'd still like a version for "green" water shots here in Puget Sound. Something that improves the viz to like 75' would really be a big seller...

Jack

Thanks, Bart! Very nice post. As a science nerd boy at heart, I loved your analysis and presentation.

All the best, James

Hi Jack,

You may also be interested in a warming filter for your West coast dives. I've heared the Tiffen 812 raises the local temperature to 81.2 degrees. I think that is Fahrenheit but I may be wrong.

Bart

This is interesting information Bart. The higher green range transmission compared to the blue range showed up when I shot in turbid (more greenish) water as these images required stronger tint correction towards magenta. However, my experience so far is that those tint corrections didn’t degrade image quality like strong kelvin corrections would do. Having no laboratory, I just shot with a 5000K light box and opened histogram for viewing content of RGB channels to get an idea. So it is interesting to see your spectral measurements. I think it is quiet useful to see what’s going on if you want to extend filter usage beyond magic’s wide operating range (deeper / greener) or considering filter choices for strobe mounting (just got a few different rolls). On the other hand, filter combinations for greater depth and more greenish water would most probably not work great just below the surface, especially in true cyan water. However I hope to find some time to try on Seychelles (not so great visibility from what I’ve read).

Your optimized filter spectral diagram comes very close to a tricolour filter. I agree that such a filter would have a strong influence on remaining light and set exposure into a totally different world. Pushing filter photography to deep limits, with a good high ISO performer and a very fast prime – would be interesting to see results and if/how far it works.

Julian

Hi Julien,

I was thinking of getting a magenta filter (or two of different strength) and take measurements and field tests of them in combination with the Magick filter. One of the tests I want to do in Cuba is to do an auto white balance on my white slate and then take pictures of the slate every few meters as I decend to 20m or so. By calculating the relative effects on the red, green, and blue channels as function of depth I should be able to get the true absorption response per channel.

If those results match the simulations than a magick filter/magenta filter of the proper relative strength should get close to ideal and probably would add a few more meters to the working depth for filters.

I looked at some tri-color filters but so far no filter seems to be quite right, although many will be significantly better than no filter. I am looking at alternative technologies that should be able to get much closer to ideal behaviour (apart from price) but want to see how things work out in the real world before discussing other ideas with Alex.

Bart

Hi guys,

If you add more magenta (or red) to the Magic filter you will increase the working depth a bit, but not much as there is so little red light down at 15-20m that there stronger filtering does very little to re-balance the spectrum.

Also by adding too much M or R the filter will no longer work close to surface (no use when that whale shark swims by). In fact if you add 40-50CC more of magenta to a magic filter, then it really wont work above 5m (you just can't get rid of the tint) - and the shots cannot be saved.

Also the extra absorbtion of the filter will cost you most of another stop of light, which will have other issues on your image quality.

I made nearly all these tests (and more) when developing the Magic Filter. So I mention this to save you wasting your time. Some of those on the WP Bahamas trip would have seen the large selection of filters I have (5, 10, 20, 40 and 50 CC of M, R, Y, C etc). The Magic filter was designed after both theoretical analysis of absorption of water, the capabilities of white balance control of both the camera and the Raw Converter and then a lot of field testing to find the most versatile filter in real world conditions while maintaining high image quality.

Of course I would encourage you to experiment with sandwich filters added to the Magic Filter. Certainly if you know that your subject will be between 12-16m, then it may be worth adding some M or R to the Magic. But this will cost you light and mean the filter won't work close to the surface.

I am just trying to save you wasting time basically re-doing the same experiments I did when developing the filter - when you could be spending that time taking great pictures with the filter!

As an example, here is a Magic Filter shot taken at exactly 15m:

Alex

Hi Alex,

I think you'd agree that most of the magic is in the camera's AWB that can correct for a considerable skew in the light spectrum. The filters just need to get in the right ball park to let the camera do the rest. The place where I am going has a shoreline dropping to about 4m, followed by a wall dropping down to 8m. I can probably dive without any filter for the first 4m and the magic filter will be in its sweet spot along the wall. No point adding any extra filters here for all my snorkeling. However, all shallow dive sites are in the 8-14m range. Based on your last picture the magic filters still help out a lot but here it would become of interest to see if some extra magenta filtering helps. I don't care if I lose the ability to shoot in the first 5m as I'm not aware of whale sharks very close to the North coast of Cuba. In the end I would not be surprised to find that the extra light loss of a magenta filter is more deleterious than any gain in spectral balance. However, it will still be fun to do the experiment.

Bart

I completely agree it is worth experimenting, and I have done those very experiments.

But I think that great images come from fussing about what your camera is pointed at, rather than the exact filter you have in the lens. That is the Magic Philosophy. Here is a solution that works. Stick it on and go and concentrate on your photography.

From my tests making the filter stronger doesn't actually add much to the working depth - as this is limited by the almost total lack or red light. And shallower than 15m the magic filter has proved to be capable of excellent image quality, with many images taken by users (not including me) now published in books, magazines, used in advertising campaigns and even Rand's David Doubilet Award at BTS.

Anyway I would suggest that you take a 20M and a 40M to try with the filter. Sandwiching filters results in slightly more light loss than if the filters are combined and made as one. Also if you sandwich more than three filters together you will see noticeable degradation to image sharpness.

One area of interest in creating a stronger filter is for cameras such as the 5D, which have fantastic high ISO performance. This camera might be able to make use of a stronger filter - although filter photography will still be limited to the top 15m because of the lack of red below that depth.

I will certainly be interested in seeing your images.

Alex

Hi Guys,

I looked at some tri-color filters but so far no filter seems to be quite right

<{POST_SNAPBACK}>

Following your theoretical approach counteracting the seawater effect entirely in greater depth would result in a filter with a very hard/steep edge around 580nm close to zero transmission below that range. Pretty similar to those kinds of harsh bandpass filters used in b/w photography. I’ve used such filters with film. Tiffen for example has some different sorts of such strong filters in different flavours which come kind of close what you are describing for greater depth (32m). I don’t recommend one of those filters to you and I honestly doubt they could succeed in any way. I just have mentioned them to express in what kind of extreme filters this approach would result for great depth like 32m (not like those black/white photography filters, but quiet close to that kind of strong bandwidth cut effect).

Even when going deeper to the limits and doing some experiments, I would go with temperature converters as starting point for further additions (That is what I prefer). I don’t expect much usage from a filter combination designed to entirely counteract the seawater in 30m. Especially the CCD diagram you have posted shows that green wavelength is still feeding the red channel a bit, maybe this has to be considered as red almost disappears? But I would like to see your results.

One area of interest in creating a stronger filter is for cameras such as the 5D, which have fantastic high ISO performance.

<{POST_SNAPBACK}>

That is definitely an interesting point to me in the future! :-)

Those have been shot with the rusty 10D + a very strong filter combination and I just can imagine/hope how much better the 5D would perform with high ISO and still quiet strong WB in such depth. But no chance at the surface this way and there is simply a limit without red in the depth (even for the 5D ).

28m:

25m:

Julian

Hi Julian,

I simulated the 32m depth curve because the computer will do whatever you ask it but I never intended to imply that filters would be practical at such depth. That's why I'm flabbergasted with your shots. Are you sure those depths are in meters not feet? They are great shots and the rocks even show real red. Shots of such large schools are great for filters as strobes can't reach that far.

Anyway, I just received my Ike housing, ports, and DS50 flash yesterday. I plan to do much of the filter test and picture taking with a 35mm f2.0 lens which should let in plenty of light if need be. My biggest problem however is going to be getting plenty "quality time" with my new toys without alienating me from my wife and family that are coming along

Bart

Do these filters only work with digital cameras og will they work with regular SLR?

Is using a filter better than correcting the colours in PS?

On film I would suggest using a UR Pro CY filter. Although you will get better results with hi-quality print film than with slide.

On film, in my experience, the UR Pro CY filter works best between 6 and 8 metres. Shallower or deeper than this it is not so pleasing. To red shallower and too blue deeper.

Alex

The link below, scroll down on the page to see 0 to 30 m shows a series of images of a Macbeth (color standard) card in clear water near noon.

http://cephschool.thecephalopodpage.org/Li...andQuantity.pdf

I have not tried Alex's filters but have heard many good things about them. I personally need to get my head around the concept of blocking light since the air/water surface and depth already are cutting into what I have to work with. For most shots, I usually want as much DoF, shutter speed and color/low graininess as I can get. However, it does seem like the ASA/ISO (sensitivity) is often the most forgiving of the three things that control the physics of exposure.

I'm just thinking out loud; a similar series with a known photographic standard like a Macbeth card or even a gray card (there are water proof ones) might be a better way to test models and show what is really happening.