I did a lot of testing of various red filters and blue/green gels back in 2022 to create a set-up suitable for mixed light video (though concepts are equally applicable to photos). This is the old thread on Wetpixel. Copying my findings here for posterity. 

 

Technique

 

A good starting point for understanding the technique is this video from Florian Fischer: https://www.facebook.com/keldanlights/videos/359120794910357/

 

How I do it:

1. With the blue filters, the color spectrum output by the video lights more or less matches the color spectrum of the ambient light at a certain depth. What that exact depth is will vary from location to location based on water clarity, time of day, cloudy/not cloudy, etc. But it will usually be somewhere in the 6-12m depth range.
2. I go down to 12 meters and do a MWB off a grey or white card using just ambient light. If the camera can't execute a good MWB at that depth without a red filter, then I use a red filter. 
3. I then lock in that MWB and turn on the video lights with the blue filters in place. There should be no/very minimal color cast to the areas of the scene that the video lights illuminate.
4. I keep the same MWB for any shots at 8m depth or deeper, down to maybe 30m. 
5. Below 30m, I just remove the blue filters from the lights and set a CWB on the camera to 5000k (the color temperature of the video lights). 
6. If I need to shoot above 8m depth, I do a MWB again at that depth based on ambient light alone. I mostly shoot with just ambient light at 8m or shallower. You could try to fill in shadows with your lights, but to avoid red color casts on foreground objects, you'll either need a 2nd set of weaker blue filters that are formulated to match ambient light at shallower depths or you'll need to use your lights at a weaker setting so the red wavelengths they put out aren't so overpowering.

 

(on lens) Red Filter Tests

 

As a bonus, I also tested the 4 orange filters I have access to against the Lee 353 + Rosco 4330 combination to see how much of the blue filter they were able to neutralize:

• No blue filter, no orange filter = 4700K, +20M
• Lee 353 + Rosco 4330 = 23000k + 121M
• Lee 353 + Rosco 4330 + Original Magic Filter = 8600k + 150M, -1.3 stops of light
  • (Worst of the bunch.. looks more yellow than orange. introduced MORE green into the image, to the point where even after doing WB in post, there was a greenish tinge. Did a decent job of countering the blue, but would not recommend using it in anything but the most pristine blue waters) 
• Lee 353 + Rosco 4330 + Auto-Magic Filter = 10000k + 121M, -1.6 stops of light
  • (Decent. Knocked out most of the blue cast, but did little for the green. Again, use only in pristine blue water)
• Lee 353 + Rosco 4330 + UR/PRO CY Filter = 7700k + 77M, -2.3 stops of light
  • (Most effective. Knocked out most of the blue AND green cast. Would be a good all-around choice)
• Lee 353 + Rosco 4330 + Keldan SF-2 Filter = 12000k + 114M, -2.5 stops light 
  • (Ok, but a bit weak. This is designed to compensate only for 4 meters of water depth, and reduces both the blue and green tint. However, in this case, the lights with the Lee 353 and Rosco 4330 filter appear to be better matched to 12 meters of depth..)

 

 

(on light source) Blue - Ambient Light - Filter Tests

 

 

1. The LEE CTB filters aren't as good as I hoped. They increase the color temperture (I.e. introduce a lot of blue) but don't introduce enough green into the light to match underwater ambient light. Results:

• No Filter = 4750K, +14 Magenta, 0 stops light lost
• Half CTB = 7550K, +6 Magenta, 2/3 stops light lost
• Full CTB = 21000K, +7 Magenta, 1 1/3 stops light lost
• 2x Full CTB = 50000K +12 Magenta, 2 stops light lost (in fact, 50000K was not sufficient to restore neutral color -- I had to dial down blue saturation to -100 in Lightroom)

 

 

2. From the Rosco filters, I tried the Cyan 30 (#4330), Cyan 60 (#4360), Cyan 90 (#4390) and Storaro Cyan (#2005). Interstingly, the Cyan 30 perfectly matched the DivePro F01C filter. I'm guessing this would be good at around 3-5m depth to match ambient light?

• No Filter = 4750K, +14 Magenta, 0 stops light lost
• Cyan 30/DivePro F01C = 6600k , +67 Magenta, 1/3 stops light lost
• Cyan 60 = 8600k, +116 Magneta, 2/3 stops light lost
• Cyan 90 = 12250K, +150 Magneta, 1 1/3 stops light lost
• Storaro Cyan = 13250K, +150 Magneta, 2 stops light lost (in fact, +150 magneta was not sufficient to restore neutral color. I had to dial down Aqua saturation -100 and Orange saturation -100 in Lightroom to get rid of the remaining color cast)

 

Based on above testing, I think the Cyan 60 might be a pretty good start for matching ambient light at 6-9 meters. Cyan 90 for 9-12 meters. Not really worth using the Storaro Cyan or the 2x Full CTB.

 

• No Filter = 4850K, +23 Magenta, 0 stops light lost 
• Rosco 4390 (Cyan 90) Filter = 8900k, +141 Magenta, 1 and 1/3 stop of light lost
• Lee 353 Filter (Lighter Blue) = 15250k, +111 Magenta, 1 stop of light lost
• Lee 724 Filter (Ocean Blue) = 26000k, +97 Magenta, 1 and 1/3 stop of light lost
• Lee 172 Filter (Lagoon Blue) = 50000k, +148 Magenta,  -100 Aqua saturation, 2.4 stop of light lost

 

I finally got my hands on the Keldan SF-6B and SF-12B blue water filters. Interesting results... (on the Canon R5 C in lightroom)

• No Filter = 5150K, +21 Magenta, 0 stops light lost
• Keldan SF-6B = 17000K, +146 Magenta,  about 1 and 2/3 stops light lost
• Keldan SF-12B = 20000K, +144 Magenta, about 1 and 2/3 stops light lost

 

Surprisingly, there wasn't a lot of difference between the SF-6B and SF-12B.. also, they were both less blue and more green than I expected.

Closest you could probably come to this would be to stack the Lee 353 filter with a Rosco 4330 or 4360 filter. 

• Lee 353 + Rosco 4330 = 23000k + 121 Magenta, about 1 stop light lost
• Lee 353 + Rosco 4360 = 27000k + 144 Magenta, about 1.5 stops light lost

 

If you wanted something closer to the SF-6B, Cyan 60 (Rosco 4360) + Lee Half CTB or Quarter CTB filter may be another combination worth trying.

 

Got a 1/2 CTB to test the earlier theory. Here's the results (on the Canon R5 C in Lightroom)

• No Filter = 4750K, +14 Magenta, 0 stops light lost
• Keldan SF-6B = 17000K, +146 Magenta,  about 1 and 2/3 stops light lost
• Keldan SF-12B = 20000K, +144 Magenta, about 1 and 2/3 stops light lost
• Lee 1/2 CTB + Rosco 4330 = 9900K +74M, about 2/3 stop light lost
• Lee 1/2 CTB + Rosco 4360 = 12000K + 104M about 1 stop light lost
• Lee 1/2 CTB + Rosco 4390 = 13500K + 129M, about 1 and 1/3 stop lost
• Lee 1/2 CTB + Rosco 4330 + Rosco 4390 = 16000K +150M, about 1 and 2/3 stop lost
• Lee Full CTB + Rosco 4330 = 22000K + 75M about  1 and 1/3 stop light lost
• Lee Full CTB + Rosco 4360 = 32000K + 99M about 1 and 1/2 stop light lost
•  

Seems it's rather hard to replicate the Keldan SF-6B filters exactly, at least with this combination of CTB and Cyan filters. You either end up overshooting the blue or undershooting the cyan. Maybe I should've bought some Lee 3/4 CTB filters to get a bit closer. I have a feeling that a Lee 3/4 CTB + Rosco 4390 Cyan filter would be pretty close.

For now, I guess I'll test the Lee 1/2 CTB + Rosco + Rosco 4330 + Rosco 4390 and Lee 353 + Rosco 4330 out in the field.

 

 

Some real world feedback ] the Red Sea -- 

I tested out both the

• Lee 353 + Rosco 4330 = 23000k + 121 Magenta, about 1 stop light lost
• Lee 1/2 CTB + Rosco 4330 + Rosco 4390 = 16000K +150M, about 1 and 2/3 stop lost

At 5 meters with DivePro G18 lights. They blended it seamlessly with an ambient white balance at this depth, filling in the shadows:

 

Just ambient light:

+ 2x DivePro G18 lights w/ Lee 1/2 CTB + Rosco 4330 + Rosco 4390

+ 2x DivePro G18 lights w/ Lee 353 + Rosco 4330

+ 4k DivePro G18 lights (2 with first filter combination, 2 with second filter combination)

So as you can see both the filter options worked well, and they worked well together too. None produced weird results. However, given the stength of the ambient light, there's a noticable difference in ability to fill in shadows (look at the mask) when using 4x 18k lumen lights vs. just 2x.

From this point on, I used the Lee 353 + Rosco 4330 combination on all 4 of my lights as I thought it did a slighly better job than the more cyan Lee 1/2 CTB + Rosco 4330 + Rosco 4390 filter combination in the conditions found in the Red Sea.

I'm pleased to report that this filter combination produced great results all the way from 5m to 30m in depth. So, setting a white balance based on ambient light and filling in with the DivePro G18 lights + Lee 353 + Rosco 4330 filters on the lights produced very pleasing results in my view. Here's an example at 30 meters:

And 20 meters:

And 15 meters:

And this is how it renders skin-tones at 20 meters. Pretty good, I think!

 the first set of frame grabs at 5m are from R5 C raw video with Canon's offical Canon Log2-12(bit) to Wide DR lut applied.

The second set at various depths is directly out of camera h.265, filmed in the EOS Standard picture profile.-

A bit of color grading of the raw files can get you very nice results indeed at 10 meters:

Hi Andrei,

 

Thank you for copying here a terrific thread from WP era that probably would be irremediably lost. I pinned it.

 

I cannot see the original WP post so maybe I'm missing something: could you explain the method you used to analyze the images? What do that numbers mean? For example:

 

Quote

Lee 353 + Rosco 4330 + UR/PRO CY Filter = 7700k + 77M

 

• 7700K 
• +77M

Thank you in advance

Yes, I skipped that part from the Wetpixel thread. TLDR Methodology:

1. Illuminate a neutral white or light grey target (a color checker chart or 90% white card or 18% grey card is ideal, but you can use white painted walls in a pinch, as I originally did) with a strobe or torch of known color balance (all my tests are done with the Divepro G18 lights, which I've separately measured at 5000k) with the filter being tested applied to the light (or, if testing transmission characteristics of an on-lens filter, applied to the lens)
2. Take a (raw) picture of the target with your camera
3. Measure the white balance in camera or in post. I bring the image into Lightroom and apply the white balance picker tool to what should be a neutral white or grey area of the image and see what resultant white balance is actually required to neutralize the color cast of the filter.  That's what the measurements above describe.
4. To estimate light loss, I keep the exposure the same between the initial test with no filter and the subsequent tests with the filters. I then bring up the exposure of the images with the filter in Lightroom using the Exposure slider (which is measured in stops) to match the non-filter picture (using the histogram and spot checker to check exposure in the same spot in the image) to see how many stops underexposed the test image is.
5. To minimize impact of ambient light on the results, do these tests in a dark room 🙂

 

My testing set-up:

 

 

And resultant images for context:

• 
• No Filter = 4850K, +23 Magenta, 0 stops light lost 
• 
• Rosco 4390 (Cyan 90) Filter = 8900k, +141 Magenta, 1 and 1/3 stop of light lost
• 
• Lee 353 Filter (Lighter Blue) = 15250k, +111 Magenta, 1 stop of light lost
• 
• Lee 724 Filter (Ocean Blue) = 26000k, +97 Magenta, 1 and 1/3 stop of light lost

 

• 
• Lee 172 Filter (Lagoon Blue) = 50000k, +148 Magenta,  -100 Aqua saturation, 2.4 stop of light lost

Edited June 24, 20241 yr by DreiFish

Thank you Andrei, now everything is clear.

 

P.S.

I have not seen a video from you in a while! Are you using th R5C only for photos?

7 hours ago, Davide DB said:

Thank you Andrei, now everything is clear.

 

P.S.

I have not seen a video from you in a while! Are you using th R5C only for photos?

 

Well.. .life interfered (newborn)and I got lazy. I don't do as much diving these days and the prospect of investing the time to film a meaningful sequence and edit it is a bit daunting 🙂 So yes, mostly taking photos these days.  There's something satisfying about getting a single 'wow' shot rather than trying to splice together something that tells a story.

 

Probably I'll return to video some day, but for now enjoying wide angle photography very much.

 

On 6/21/2024 at 11:13 PM, DreiFish said:

I did a lot of testing of various red filters and blue/green gels back in 2022 to create a set-up suitable for mixed light video (though concepts are equally applicable to photos). This is the old thread on Wetpixel. Copying my findings here for posterity. 

 

Technique

 

A good starting point for understanding the technique is this video from Florian Fischer: https://www.facebook.com/keldanlights/videos/359120794910357/

 

How I do it:

1. With the blue filters, the color spectrum output by the video lights more or less matches the color spectrum of the ambient light at a certain depth. What that exact depth is will vary from location to location based on water clarity, time of day, cloudy/not cloudy, etc. But it will usually be somewhere in the 6-12m depth range.
2. I go down to 12 meters and do a MWB off a grey or white card using just ambient light. If the camera can't execute a good MWB at that depth without a red filter, then I use a red filter. 
3. I then lock in that MWB and turn on the video lights with the blue filters in place. There should be no/very minimal color cast to the areas of the scene that the video lights illuminate.
4. I keep the same MWB for any shots at 8m depth or deeper, down to maybe 30m. 
5. Below 30m, I just remove the blue filters from the lights and set a CWB on the camera to 5000k (the color temperature of the video lights). 
6. If I need to shoot above 8m depth, I do a MWB again at that depth based on ambient light alone. I mostly shoot with just ambient light at 8m or shallower. You could try to fill in shadows with your lights, but to avoid red color casts on foreground objects, you'll either need a 2nd set of weaker blue filters that are formulated to match ambient light at shallower depths or you'll need to use your lights at a weaker setting so the red wavelengths they put out aren't so overpowering.

 

(on lens) Red Filter Tests

 

As a bonus, I also tested the 4 orange filters I have access to against the Lee 353 + Rosco 4330 combination to see how much of the blue filter they were able to neutralize:

• No blue filter, no orange filter = 4700K, +20M
• Lee 353 + Rosco 4330 = 23000k + 121M
• Lee 353 + Rosco 4330 + Original Magic Filter = 8600k + 150M, -1.3 stops of light
  • (Worst of the bunch.. looks more yellow than orange. introduced MORE green into the image, to the point where even after doing WB in post, there was a greenish tinge. Did a decent job of countering the blue, but would not recommend using it in anything but the most pristine blue waters) 
• Lee 353 + Rosco 4330 + Auto-Magic Filter = 10000k + 121M, -1.6 stops of light
  • (Decent. Knocked out most of the blue cast, but did little for the green. Again, use only in pristine blue water)
• Lee 353 + Rosco 4330 + UR/PRO CY Filter = 7700k + 77M, -2.3 stops of light
  • (Most effective. Knocked out most of the blue AND green cast. Would be a good all-around choice)
• Lee 353 + Rosco 4330 + Keldan SF-2 Filter = 12000k + 114M, -2.5 stops light 
  • (Ok, but a bit weak. This is designed to compensate only for 4 meters of water depth, and reduces both the blue and green tint. However, in this case, the lights with the Lee 353 and Rosco 4330 filter appear to be better matched to 12 meters of depth..)

 

 

(on light source) Blue - Ambient Light - Filter Tests

 

 

1. The LEE CTB filters aren't as good as I hoped. They increase the color temperture (I.e. introduce a lot of blue) but don't introduce enough green into the light to match underwater ambient light. Results:

• No Filter = 4750K, +14 Magenta, 0 stops light lost
• Half CTB = 7550K, +6 Magenta, 2/3 stops light lost
• Full CTB = 21000K, +7 Magenta, 1 1/3 stops light lost
• 2x Full CTB = 50000K +12 Magenta, 2 stops light lost (in fact, 50000K was not sufficient to restore neutral color -- I had to dial down blue saturation to -100 in Lightroom)

 

 

2. From the Rosco filters, I tried the Cyan 30 (#4330), Cyan 60 (#4360), Cyan 90 (#4390) and Storaro Cyan (#2005). Interstingly, the Cyan 30 perfectly matched the DivePro F01C filter. I'm guessing this would be good at around 3-5m depth to match ambient light?

• No Filter = 4750K, +14 Magenta, 0 stops light lost
• Cyan 30/DivePro F01C = 6600k , +67 Magenta, 1/3 stops light lost
• Cyan 60 = 8600k, +116 Magneta, 2/3 stops light lost
• Cyan 90 = 12250K, +150 Magneta, 1 1/3 stops light lost
• Storaro Cyan = 13250K, +150 Magneta, 2 stops light lost (in fact, +150 magneta was not sufficient to restore neutral color. I had to dial down Aqua saturation -100 and Orange saturation -100 in Lightroom to get rid of the remaining color cast)

 

Based on above testing, I think the Cyan 60 might be a pretty good start for matching ambient light at 6-9 meters. Cyan 90 for 9-12 meters. Not really worth using the Storaro Cyan or the 2x Full CTB.

 

• No Filter = 4850K, +23 Magenta, 0 stops light lost 
• Rosco 4390 (Cyan 90) Filter = 8900k, +141 Magenta, 1 and 1/3 stop of light lost
• Lee 353 Filter (Lighter Blue) = 15250k, +111 Magenta, 1 stop of light lost
• Lee 724 Filter (Ocean Blue) = 26000k, +97 Magenta, 1 and 1/3 stop of light lost
• Lee 172 Filter (Lagoon Blue) = 50000k, +148 Magenta,  -100 Aqua saturation, 2.4 stop of light lost

 

I finally got my hands on the Keldan SF-6B and SF-12B blue water filters. Interesting results... (on the Canon R5 C in lightroom)

• No Filter = 5150K, +21 Magenta, 0 stops light lost
• Keldan SF-6B = 17000K, +146 Magenta,  about 1 and 2/3 stops light lost
• Keldan SF-12B = 20000K, +144 Magenta, about 1 and 2/3 stops light lost

 

Surprisingly, there wasn't a lot of difference between the SF-6B and SF-12B.. also, they were both less blue and more green than I expected.

Closest you could probably come to this would be to stack the Lee 353 filter with a Rosco 4330 or 4360 filter. 

• Lee 353 + Rosco 4330 = 23000k + 121 Magenta, about 1 stop light lost
• Lee 353 + Rosco 4360 = 27000k + 144 Magenta, about 1.5 stops light lost

 

If you wanted something closer to the SF-6B, Cyan 60 (Rosco 4360) + Lee Half CTB or Quarter CTB filter may be another combination worth trying.

 

Got a 1/2 CTB to test the earlier theory. Here's the results (on the Canon R5 C in Lightroom)

• No Filter = 4750K, +14 Magenta, 0 stops light lost
• Keldan SF-6B = 17000K, +146 Magenta,  about 1 and 2/3 stops light lost
• Keldan SF-12B = 20000K, +144 Magenta, about 1 and 2/3 stops light lost
• Lee 1/2 CTB + Rosco 4330 = 9900K +74M, about 2/3 stop light lost
• Lee 1/2 CTB + Rosco 4360 = 12000K + 104M about 1 stop light lost
• Lee 1/2 CTB + Rosco 4390 = 13500K + 129M, about 1 and 1/3 stop lost
• Lee 1/2 CTB + Rosco 4330 + Rosco 4390 = 16000K +150M, about 1 and 2/3 stop lost
• Lee Full CTB + Rosco 4330 = 22000K + 75M about  1 and 1/3 stop light lost
• Lee Full CTB + Rosco 4360 = 32000K + 99M about 1 and 1/2 stop light lost
•  

Seems it's rather hard to replicate the Keldan SF-6B filters exactly, at least with this combination of CTB and Cyan filters. You either end up overshooting the blue or undershooting the cyan. Maybe I should've bought some Lee 3/4 CTB filters to get a bit closer. I have a feeling that a Lee 3/4 CTB + Rosco 4390 Cyan filter would be pretty close.

For now, I guess I'll test the Lee 1/2 CTB + Rosco + Rosco 4330 + Rosco 4390 and Lee 353 + Rosco 4330 out in the field.

 

 

Some real world feedback ] the Red Sea -- 

I tested out both the

• Lee 353 + Rosco 4330 = 23000k + 121 Magenta, about 1 stop light lost
• Lee 1/2 CTB + Rosco 4330 + Rosco 4390 = 16000K +150M, about 1 and 2/3 stop lost

At 5 meters with DivePro G18 lights. They blended it seamlessly with an ambient white balance at this depth, filling in the shadows:

 

Just ambient light:

+ 2x DivePro G18 lights w/ Lee 1/2 CTB + Rosco 4330 + Rosco 4390

+ 2x DivePro G18 lights w/ Lee 353 + Rosco 4330

+ 4k DivePro G18 lights (2 with first filter combination, 2 with second filter combination)

So as you can see both the filter options worked well, and they worked well together too. None produced weird results. However, given the stength of the ambient light, there's a noticable difference in ability to fill in shadows (look at the mask) when using 4x 18k lumen lights vs. just 2x.

From this point on, I used the Lee 353 + Rosco 4330 combination on all 4 of my lights as I thought it did a slighly better job than the more cyan Lee 1/2 CTB + Rosco 4330 + Rosco 4390 filter combination in the conditions found in the Red Sea.

I'm pleased to report that this filter combination produced great results all the way from 5m to 30m in depth. So, setting a white balance based on ambient light and filling in with the DivePro G18 lights + Lee 353 + Rosco 4330 filters on the lights produced very pleasing results in my view. Here's an example at 30 meters:

And 20 meters:

And 15 meters:

And this is how it renders skin-tones at 20 meters. Pretty good, I think!

 

Well done, DreiFish!! I used a Full CTB  and Keldan Orange filter on the Olympus, yet on Sony I juslt almost of the time a natural light with WB each 3 meters! I will try in a pool with Sony. I note with filters I lost a lot of light(my ones are from China, yet still working for 3 years - i just spend 85€ each on Amazon - LetonPower)
Thanks for sharing!!

Big hug from Portugal

As an update, I ran a few more tests with the Keldan Spectrum filter + DIY ambient light filters, mostly the Rosco 4360 and Rosco 4390, in the slightly bluer waters of Okinawa main island.
The 4360 was ok, but did let too much reds through, 4390 less conclusive...

I finally bit the bullet and bought sheets of the above Dreifish formula: a Rosco 4330 + Lee 353 filter gel combo

Tested it this morning at a rather murky site but results were excellent (I shoot with a Keldan SF-1.5 gel on the lens, as a Panasonic manual WB does not do great without one, leaving a cast).

With the ambien filters on the lights, I could move from full ambient wide to closeup shots with no red-peak, film under corals and under overhangs without changing the white balance.
Waaaaay better than a single Rosco, and just hits the right notes spectrum-wise, neutralising the red spike.


Even with my limited lighting (2x Kraken Hydras 8000lm and 1x Backscatter MW 4000lm),it works fine the setup gives you highlights, details, and lets you film where the sun don't shine...
So everything I wanted...

¥Having access to "Portable sunlight" would be a way to describe it.

This was a quick-and-dirty test - I'll post some clips when I can test in better conditions, with a wide-lens instead of naked port.

Until then, Long Live The DreiFilter! 🤩

Here are some unedited video screen caps, straight out of camera:
Parameters: Keldan SF -1.5 on the Lumix LX10 flat port, Dreifilter on my 3 lights, white balance to ambient light with a Whibal card, 10m depth, ISO 400, f/6.3, lights at 80% power to get highlights and fills.

moving out (close-up to wide scene - note: the bubble anemone is bleached out, and white as it should be)

peeking under a coral, with an ambient light WB!



ye ol' mask shadow test


3 Dreifilters + Keldan SF-1.5 (DIY filter mounts using step-up / step down adapters and an o-ring)


cheers!

Ben

Edited October 5Oct 5 by bghazzal