The GetDPI Photography Forum

Great to see you here. Join our insightful photographic forum today and start tapping into a huge wealth of photographic knowledge. Completing our simple registration process will allow you to gain access to exclusive content, add your own topics and posts, share your work and connect with other members through your own private inbox! And don’t forget to say hi!

Balancing foveon colour channels - which filter?

I'm always years behind everyone else technologically :eek:, and have just been reading about balancing the sensitivity of the colour channels before the sensor to improve overall dynamic range.
From my simple (but possibly completely wrong) calculations, I am speculating that adding an orangish filter on the lens should bring the channels closer together - my guess is either an 85C or an 81EF (I think reality is somewhere inbetween).
Has anyone here done any experiments (or REAL calculations!) to find out which value of filter is most appropriate for everyday use?

...or is it a complete waste of time due to variable external factors?
 

darr

Well-known member
The filters you mention are warming filters. From my film days, I used a 85C when I had to shoot tungsten film in daylight. Can you not balance the colors in post-processing?
 

adsf

New member
I tried a B+W KR3. It makes everything warmer which i like. Its not greenish anymore but still blotchy.
 
Can you not balance the colors in post-processing?
It's not about the colour as an end result, but about what hits the sensor.

Out of the box, with a foveon sensor, the blue channel is more sensitive than the green, and the green is more sensitive than the red.
If I understand it correctly, the idea is to put a coloured filter on the lens which balances out the difference between the colour channels, so that, for white light, each channel will clip at the same point, and similarly run into noise issues at the same point - instead of one channel clipping or having noise issues before the others, which would reduce the possible dynamic range.

The same effect is seen with bayer filters, but the channel sensitivity is different - the suggested filter for bayers seems to be a CC30M. I haven't been able to find any info anywhere on what (if any) foveon users are using.
 
I tried a B+W KR3. Its not greenish anymore but still blotchy.
I think we're still going to run into blotch problems when underexposed, but hopefully if we can gain a stop or two of useful range it will happen less often.
At least that's what I'm hoping... ;)
 

darr

Well-known member
It's not about the colour as an end result, but about what hits the sensor.

Out of the box, with a foveon sensor, the blue channel is more sensitive than the green, and the green is more sensitive than the red.
If I understand it correctly, the idea is to put a coloured filter on the lens which balances out the difference between the colour channels, so that, for white light, each channel will clip at the same point, and similarly run into noise issues at the same point - instead of one channel clipping or having noise issues before the others, which would reduce the possible dynamic range.

The same effect is seen with bayer filters, but the channel sensitivity is different - the suggested filter for bayers seems to be a CC30M. I haven't been able to find any info anywhere on what (if any) foveon users are using.
I think of the Foveon technology similar to a different type of film (i.e., B/W panchromatic, orthochromatic, isochromatic, etc.) and the reason it has a particular look. I think changing the color channels may change the Foveon look. Have you tried in depth testing with ETTR exposure to help improve noise issues that may help with DR?
 

darr

Well-known member
The belief is that using ETTR along with balanced colour channels helps you to make the most of what the camera can do.
I get excellent results using ETTR and bracketing. Sometimes art can be found in the flaws. ;)
 
I managed to beg, borrow (but not steal) a few filters to play with in order to give this theory a try, and the results unfortunately are not good.

First of all, my filter guess was not even close. I ended up putting an 85 and an 85B together... and it was still nowhere near.

By my (quick) tests:

Unfiltered, 1.9 stops between the blue channel and the red.
85 + 85B, 1.0 stops between the blue and red channels.

Of course with the filter factor you are losing 1.3 stops.
Essentially, then, this technique (as far as foveons go) makes your noise WORSE. You are balancing the channels, but the light lost due to the filters negates any dynamic range you have gained - and in fact your noise will be starting on all three channels instead of just one.

At least that's how it looks to me... :cry:
 
I've had a rethink about what I said in my last post, and I'm not sure I agree with myself...! :D

If the issue for a particular shot is one of noise with limited light, then what I said previously still stands. If, however, there is plenty of available light (or room to manouvre on the exposure itself) then using the above technique and ETTR should give you extra dynamic range before hitting the noise floor, as long as the longer exposure isn't adding to the noise problem.

Phew!

That leaves me back with the original problem though - is there a filter strong enough, and which one is it?
 
I managed to get hold of some stronger filters to push this trial as close as I could to what was expected, and it turns out it's not a valid option (not with my results, anyway).

The results highlight an issue I was made aware of on another forum; that of a poor performance of the Foveon with regards to colour separation. You CAN get each channel to produce similar response to filtered white light, but by that point you've lost any hope of the colours recorded being remotely accurate.

I also hit the same issue trying to find UniWB. Fruitless.

I think Darr's bracketing approach is probably the only sensible one!
 

xpatUSA

Member
I managed to get hold of some stronger filters to push this trial as close as I could to what was expected, and it turns out it's not a valid option (not with my results, anyway).

The results highlight an issue I was made aware of on another forum; that of a poor performance of the Foveon with regards to colour separation. You CAN get each channel to produce similar response to filtered white light, but by that point you've lost any hope of the colours recorded being remotely accurate.
I would recommend buying the research edition of RawDigger. Then you will be able to open X3Fs and see a 3-channel histogram of the raw data in the image or part thereof. Here's a SD14 white card shot:



The x-axis is in EV. The SD14 raw data saturates up around 8000, so I could have gone another 1/3 EV in the shot. (Please don't tell me about the 12-bit ADC, Sigma does quite a bit of scaling in-camera and the raw data is written in unsigned 16-bit format.)

Notice that the sensor channel differences are not that bad when you consider that the lamp is a domestic CFL and, of course, that the raw data is not in RGB color space and neither is it white balanced which is why 'green' is a little higher than the other two channels.

So, your concerns seem to be more about SPP or whatever converter you use. The conversion transforms are of necessity extreme compared to Bayer.

I also hit the same issue trying to find UniWB. Fruitless.
I used UniWB as the custom WB on a SD10 for a long while. I followed Luijck's instructions to the letter, took the screen shot and it worked a bit but not enough for my shots to show that much difference.

I think Darr's bracketing approach is probably the only sensible one!
Indeed, bracketing works and is easy to do.

As to improving DR, the sad part is that the Foveon blotching is intrinsic to the technology. On the F19 (Polaroid x530) sensor it can even occur at mid-tones, giving a DR of, like, 1EV :( . With later sensors and SPP making it's best efforts, you might get as much as a 4EV shot . . . or you might not. Of course, Foveon blotching isn't really "noise" as we understand it, more like artefacts I suppose. And, as I've said elsewhere, SPP is the finest de-blotcher on the planet, whereas DCraw is definitely not ;)

best regards,

Ted
 
Last edited:
Top