Public Lab Research note


Focus on filters

by cfastie | November 07, 2013 16:28 | 5,565 views | 11 comments | #9744 | 5,565 views | 11 comments | #9744 07 Nov 16:28

Read more: stable.publiclab.org/n/9744


Ned Horning has decided that the Canon Powershot A2200 is a good choice for kite, balloon, and UAS aerial photography. He watches the Canon site where refurbished ones sell for $70 but sometimes are on sale for $55. He had two A2200s with broken ribbon cables, so I ordered replacements on eBay ($11 for both including shipping) and figured out how to install them.

The A2200 seems to be a good replacement for the A495 family. It is smaller and lighter because it uses a Li-ion battery instead of two AAs. I like AAs better, because I invested in lots of AA Eneloops, but they do have a weight penalty. The A2200 is newer and CHDK can be installed on SD cards larger than 4 GB without partitioning the card. That is a major convenience. The A2200 is a little smarter because it has the newer Digic IV processor, but the two cameras have very similar features. For example they both have custom white balance and infinity focus mode but lack manual focus, shutter priority, and image stabilization. The A2200 has 14 MP instead of 10 MP, but 10 MP is all you need for aerial photography with these little lenses and sensors. They both have 1/2.3" sensors like lots of Powershots. I purchased a glass BG3 (Superblue) filter for $16 and have been looking for a camera to install it in. It's the proper size for these cameras.

The A2200 is pretty easy to convert for infrared photography (unlike the A2400). It's much easier if you unplug a ribbon cable, which requires flipping up a little brown clamp which loosens the connection. The broken ribbon cables were on the back button pads, and replacing them required some extra disassembly (one deep screw to remove the top control assembly) and learning how to open a second type of ribbon cable connector (the black plastic cover hinges up). I installed the BG3 glass filter in one camera and a piece of gelatin Wratten 25A filter in the other. Ned is promoting the red filter as superior to blue filters for NDVI photography.
.

DamFocusBG3_2k.jpg
Infrablue photo from A2200 with BG3 glass filter inside. Image clarity seems to be better than with polyester (Rosco) infrablue filters, but focus is poor around the edges of the photo. This is consistent among many test photos.
.

I was expecting great images with the glass BG3 filter installed, but I am a little disappointed. Although the glass filter should preserve the optics of the original IR block filter, focus does not seem to be as good as with an unmodified camera. Focus is poor at the edges of the photos. The severity of the blur varies, but is present in many test photos using different f/stops and focal lengths. The same problem is present in photos from the other camera with the Wratten 25A gelatin filter.
.

house2k.jpg
Details of two photos taken by a Canon A2200 with Wratten 25A gelatin filter installed internally. The top photo was taken using Program mode and the blue color reflects the custom white balance setting. The lower photo was taken in Auto mode and the color reflects the white balance preset chosen by the camera. The left halves of both photos have been desaturated to better compare image clarity. The photo taken in Program mode is crisper than the one in Auto mode. Both photos are crisper in their centers than near their edges.
.

There is a current discussion at the plots-infrared Google Group about camera focus after infrared conversion. Kempo noted that a converted A480 had better focus in Auto mode than in Program mode. I was intrigued, so I tried it. When in Auto mode, shutter speed, f/stop, white balance, and who knows what else are selected by the camera. So you cannot use a custom white balance in Auto mode which makes it unhelpful for infrablue photography. Unlike Kempo, I found that the photos I took in Program mode were crisper than the ones in Auto mode. but they were both more blurry near the edges.

I am not sure if there is anything that can be done to improve the crispness of these photos. The problem could be related to the optics of infrared light compared to visible, so I should examine the separate channels in each photo, some of which are all NIR light and some of which are mostly visible. The inconsistent results with Auto and Program mode are puzzling. I don't know why either mode would degrade clarity, although when the camera chooses f/stop, shutter speed, ISO, and white balance those settings could differ dramatically from Program mode settings, and each of those settings could influence clarity. More tests are needed.


11 Comments

Very interesting line of inquiry. I've always been a bit confused about the role of the filter material itself in focus -- my assumption has been that the focus algorithm may be "confused" by monochromatic images; i.e. it may use a combination of channels to calculate focus, and have trouble focusing if one or more of those channels has very little light.

But I hear people talking about filters having physical optical effects, and you mention:

the glass filter should preserve the optics of the original IR block filter

If the filters are not curved at all, how can this be? They also don't change the geometry of the existing lens/sensor arrangement; and since it's rubber-gasketed in place, I had guessed that the precision of its placement was not very important.

Chromatic aberration seems like a reasonable possibility though -- especially with infrablue, where the span of wavelengths (blue vs. IR) is quite wide. With Ned getting good red-filter infragrams, the two channels would be much closer in wavelength, and maybe the lower chromatic difference would make for better overall focus. But even in infrablue images, we should be able to see some difference in focus between the blue and IR channels, unless the focus algorithm just focuses quite evenly halfway between the two channels, in which case they'd be equally blurry.

Is this a question? Click here to post it to the Questions page.

Reply to this comment...


I was assuming that the best way to preserve image quality when removing a glass filter would be to replace it with another glass filter. Plastic, gelatin, or film filters are not the same thickness, do not have optically smooth surfaces, may not be flat, and may not be internally homogeneous (film and sometimes plastic filters have layers). All those things could influence image clarity. I was also very impressed with the quality of the Superblue photos from Ned's Canon G11 with a glass BG3 filter inside, and was naively hoping to match that with the small sensor and lens of the A2200.

There is some chromatic aberration evident in some of these photos -- a faint blue ghost on the outside edge of objects near the edge of Wratten 25A photos. Further investigation is warranted.

Reply to this comment...


The filter should have nothing to do with the difference in quality between the center and edge of the image since it is uniform in density. Lenses, by their nature do not focus in a plane and thus this is one of the common quality measures for lenses. F stop with film cameras effects depth of field. Just like a pin hole camera, if the aperture is small enough the depth of field is infinite. Thus if we had an infrared sensor with a pin hole for the lens it would be in focus automatically. I do not know if these digital cameras that we are using physically change the aperture or algorithmically accomplish the equivalent. What is needed is to determine 1) the focus algorithm, and 2) the aperture mechanism.

Reply to this comment...


Have you considered watching other sites for cameras that are refurbished? One I would recommend is canonpricewatch.com. It also shows prices for other point and shoots if you might want to look at another one. Just an idea to try and help out.

Is this a question? Click here to post it to the Questions page.

Reply to this comment...


The aperture mechanism definitely varies among Powershots. By peering into the front of the lens I learned that the high end ones have an actual iris with leaves and can produce an aperture of multiple sizes. The A495 apparently has only two different apertures which seem to be round but I don't know how one replaces the other (elastic?). The simpler cameras have no variable aperture and instead move one or more neutral density filters into the optical path. So f-stop does not influence depth of field in the SD1100 and A2200.

......Camera...... .............. Variable Iris.......... .......ND filter(s)
SD1100 no yes
A2200 no yes
A495 yes (2 round holes) ?
SX120 yes (~6 leaves) ?
S95 yes (6 leaves) ?

.
Okay, the markdown above displays properly in preview mode, but not when published. So...

Camera --- Variable Iris--------ND filter(s)

SD1100------no----------------yes
A2200--------no----------------yes
A495----------yes (2 round holes)---?
SX120--------yes (~6 leaves)-------- ?
S95------------yes (6 leaves)-------- ?
.

Assuming the BG3 glass filter I installed is seated properly, the soft edges in the A2200 photos must be due to different optical properties of the glass compared to the original IR block filter (different thickness or glass density) or a failure to properly focus NIR light compared to visible, or both. First I should open up the camera again and wiggle everything to see if that helps.

There are lots of A2200 cameras for sale for $70, and some for $55, but those don't include shipping. The Canon refurbished store sells them for $70 including shipping, and they have periodic sales. For this week, you get $20 off if you spend $100.

Is this a question? Click here to post it to the Questions page.

Reply to this comment...


The Canon A2200 (refurbished) is on sale today (November 21) for $48.99 (free shipping) with a one year warranty: http://shop.usa.canon.com/shop/en/catalog/cameras/refurbished-powershot-digital-cameras/powershot-a2200-blue-refurbished (red, blue, or black).

Reply to this comment...


Couple quick thoughts on the filter/clarity issue. For starters, I just ordered an a2200 for the infra blue purposes from eBay, so I haven't had a chance to look at the ir block filter myself yet, but:

  1. if the block filter is merely coated with some dye or material to block ir, could it be possible to chemically remove the coating without damaging the filter's glass? If yes, then the ir filter could be placed back in the camera and this should resolve the focusing issue...

  2. Alternatively, I read on another thread that the camera developers build in a negative refraction factor into either the firmware and/or sensor hardware to adjust for the known optical refraction of the ir block filter, and therefore removing the ir block still leaves in place this other built-in compensation. If this is the case, and the correction is built into the firmware, perhaps there could be some form of chdk correction to undue the negative compensation?

These ideas occurred to me as I fell asleep last night forcing me to wake up and write them down so I wouldn't forget, haha.

Is this a question? Click here to post it to the Questions page.

Reply to this comment...


  1. Interesting idea. Some IR blocking filters are "hot mirrors" which are coated to reflect IR. Maybe it's possible to remove the coating (but maybe not so easy). Other IR blocking filters are absorbtive glass, so they cannot be modified.
  2. Tom Benedict started a good thread at the KAP forum on single camera IR systems and mentioned that:

    "In the case of IR blocking filters, these are fixed characteristics of the optical system at the time the lenses are designed. So most lens designers will design in a little negative spherical to compensate for the IR blocking filter. When you remove the IR blocking filter, you essentially introduce negative spherical into the optical prescription. It helps if you can put that back in. That's why a number of places that do full-spectrum conversions will stick an AR coated BK7 window in front of the detector."

He was referring to optical compensation. Correcting for image distortion in software is possible, but correcting for spherical aberration in software might be tough (how would the software know which of the photons hitting a pixel did not belong there?).

Is this a question? Click here to post it to the Questions page.

Reply to this comment...


@Cfastie,

That post from Tom was the one I was referring to in my idea #2 but yeah, I must have misread it.

Still, regarding #1, I do understand that some materials are coated and others are actually naturally absorbative, my question is if anyone knows which the A2200 filter is? As for the difficulty of removing a coating, I'd be willing to try it when my camera arrives provided it's a coated filter.

Is this a question? Click here to post it to the Questions page.

Reply to this comment...


Chris, do you remember/know the dimensions of the A2200 hot filter?

Is this a question? Click here to post it to the Questions page.

Reply to this comment...


I think most small Powershots use an IR block filter that is 8.9 x 7.9 mm. That's the advertised size of the BG3 glass filter I put into an A2200 and it fit fine. I measured the filter I removed from an old A590 and it was essentially the same: 8 x 9 mm. The BG3 filter was advertised as 0.55 mm thick, and I measured the A590 filter with cheap calipers as 0.36 mm.

I did a request for quotation at Alibaba for AR coated BK7 glass in that size and got some quotes, but they were mostly unintelligible and unhelpful. There is a piece on ebay for $49 that he will cut to size.

Is this a question? Click here to post it to the Questions page.

Reply to this comment...


Login to comment.