A year ago I did not have a clue about the limitations of my equipment. Now know them. That by itself is a big progress. I know how to use my camera appropriately and my pictures are visibly better than before.

The last piece of equipment I bought is a BluRay burner. The DVDs are neither big enough to store all the pictures that I take within a year nor fast enough to allow me to browse through them. Storing them on a hard drive works, but you have to take care of 2 potential problems – accidental deletion and hard disk failure. As all I.T. guys know the solution is called “off site backup.” Not only you need backup of your priceless pictures, but you should store the backup outside of your primary location. Now that the BD-R discs can be bough for less than $2 a piece there is a quick and easy solution for those backups. Here are 3 extra advices from me:

1. Write the BD-R(s) slowly. Both the device and the media that I bough are capable of writing at 12x speed, which is about 9 minutes for 25GB of data. I did verify, that writing at 6x instead produces a much better result. It takes less than twice longer to write the disk (I know why but it is getting too technical), but after that the disk that I wrote at the lower speed can be read about 20% faster than the other. That means only one thing – the fast writing creates problems, that the system is able to correct during the read, but the correction takes extra time. You should avoid that – those data recovery capabilities are designed to correct issues caused by the storage/aging/handling of the media and they have their limits. As those problems accumulate through the years the disk eventually becomes unreadable and the data is lost forever. Those extra 5 minutes waiting when writing at 6x instead of at 12x are going to extend the usable lifespan of the media with several years. Find where you can set the write speed and lower it from the maximum – it is worth it.
2. Always have a plan B – make 2 copies and store them separately. Your mother’s home is an excellent location for the second copy, your kid’s house can work too. You can rent a bank vault too.
3. Periodically check that your backups are still readable. The usable lifespan of BD-R media is still unknown and of course it depends on both the media used and the storage conditions. When you see the first error copy whatever you can salvage to a new media.

If you enjoyed reading my blog or if your pictures became better as a result of it feel free to post a comment here or there. I may not be around to reply, but you may help others.

If you want to read further about the technical aspect of the photography and understand the different capabilities and limitations that exist I highly recommend reading mansurovs blog. There you can learn about the benefits of full frame cameras compared to the crop sensor cameras, the focusing issues experienced by the owners of F/1.4 lenses, the different types of color aberrations and how to deal with them, the role of the filters in the digital photography and a lot more.

Two days ago I took a picture of the Christmas tree using this technique. Here is a crop of it:

I definitely enjoyed the perfect white balance I obtained and shared the picture with the family. The reaction? Something like “What are those colors? Turn around and see that the colors are different!” So I did. Sure the colors are different. Way different in fact. So, I reverted back to the built-in Tungsten white balance and produced this:

That was accepted much better by my audience and I was done for the night.

Next day I presented the 2 pictures to everyone in of my photography enthusiast buddies. They all agreed – the second picture is better, the first one looks too artificial.

Now I referred them to one of our favorite photography web site – dpreview. Here is a quote from their review of my camera, although a similar one can be found on Nikon D300 or D3S:

Artificial light White Balance

It’s getting a bit tiresome to have to say this in every Canon SLR review but the facts are pretty obvious, the EOS 450D doesn’t do automatic white balance in artificial light, full stop. If you want white whites and you’re indoors or in any mixed light situation, you will almost definitely need to take a manual preset or use the Kelvin temperature option. The whole ‘we believe photographers want a representation of the light color in the scene’ argument falls down when you consider that your eye doesn’t see the light anywhere near as yellow as the camera captures it (and in any case if that’s the intention then give the photographer the choice of ‘Accurate white AWB’ or ‘Representative AWB’ modes).

Another trusted source – cambridgeincolor – starts their Tutorial on White Balance this way:

White balance (WB) is the process of removing unrealistic color casts, so that objects which appear white in person are rendered white in your photo.

Is it possible, I wonder, that both Canon and Nikon DSLRs cannot do a proper white balance and how is that everyone seems to like the “bad” color balance produced with the default setting more than the perfect white balance achieved using the a specially designed gray card and software? Is it that “Accurate white AWB” is useless unless you will later intentionally adjust the color cast to simulate the “atmosphere”?

This is all I wanted to share today. While writing the post my daughter looked through my shoulder and ask: Are you uploading a picture of our tree on the Web? I explained the idea to her and asked which picture is better, expecting to add one more person to the already perfect consensus in favor of the second version. No such luck – she likes the first picture, because the colors on the snow man are as they are supposed to be. That is likely related that she has seen the snow man on day light and knows what the colors on it are and the yellow cast does not please her.

Which one do you prefer?

As a matter of fact my technique has improved significantly in the past year, so the yellow on my pictures is yellow just out of the camera. My eye is better trained to notice over and underexposure too. One thing that I noticed in addition to the yellow leaves on my pictures is the overexposed highlights. No, they don’t blink on the camera screen, but they clearly have blown channel or two. This reminded me that I read somewhere that after putting the CPL on your lens you should remember to dial in some negative exposure compensation, around 1/2 or 2/3 EV, I don’t remember exactly. The part that I do remember is that this is done to prevent the camera meter from undoing the effect that you are trying to achieve with the CPL. Well, I remember it, but I never understood it. So I decided to give it a try.

All the picture below are with the CPL properly oriented to maximum effect. The second image is with -2/3EV exposure compensation and the third image is produced from the same RAW file as the second image, but with the maximum shadow correction that DPP allows (which may be too much, but the objective here is to show that the shadows can actually look lighter than the shadows without the exposure compensation.)

Picture 1, no EC:

Picture 2, -2/3 EV EC – the colors of the leaves are visibly better, but the shadows are too dark:

Picture 3, same RAW as picture 2, shadows opened in DPP:

Here are a few details, cropped at 100%, in the same order as above.

As a matter of fact I did some pixel level observations, which clearly show that on the red leaves the clipping is only in the red channel, the yellow leaves have a lot of pixes with both red and green clipping and even some of the green leaves show both read and green clipping, although not that frequently.

Well, the red color clipping does not surprise me, I actually expected it. But clipping the green?! I have seen more than one reputable sources saying that blue and red clip easily because of the sensor having twice more green compared to blue and red areas. Unless the overexposure is so strong that both red and green clip, just that the reds clip more than the greens.

In my mind there is no question that the second picture retained considerably more highlight details compared to the first one. I personally prefer the modified version with the brighter highlight – it looks much better and you would not necessary question the amount of shadow details unless you have struggled with the subject yourself.

Well, I bough Rosco Strobist Collection of color gels for the flash from B&H Photo together with a gray card and started my journey towards better colors.

On the Rosco website you can find 2 PDF files with very detailed description of the matter. But life proved to be more difficult that it looks like.

The first issue is that I cannot establish a link between the custom color balance inside Canon XSi (450D) and the color temperature settings in DPP. This is the sequence:

1. Using Auto White Balance shoot the gray card (shows orange as expected)
2. Use the picture from step 1 to setup custom white balance in the camera and switch the color balance to custom white balance
3. Using the newly established custom white balance shoot the gray card again. This time it should look perfectly gray.
4. Download the file from the previous step in DPP, switch it to Color Temperature and try getting a gray color by adjusting the slider. Well, it is not possible.

Plan B was to try the same operation with ACDSee. Much better results there – on the color tab in RAW Editing this software shows you the white balance as selected by the camera. To my surprise it is represented by 2 parameters: color temperature in K and something called Tint. It was not very surprising to find the color temperature at 2670K, pretty damn close to the advertised 2700K, but what does 30 as value for Tint supposed to mean. Wikipedia is not very helpful – stuff gets very deeply mathematical very quickly. So, back to DPP. There is no Tint over there. Under Tune you can find a color wheel, but the parameters there are color tone and color saturation – it is not very obvious how that relates to the Tint value that ACDSee gave me.

The issue described above does have a workaround – just remember to do a test shoot of the gray card all the time and set the white balance based on it. That is apparently better than just shooting the gray card and setting the white balance in software, since DPP does not want to average the color of the gray card and every pixel has a different color.

Having set the proper custom white balance it is time to find the right (combination of) color gel(s) for the flash. The ROSCO’s brochure has certain recommendations about that, but they don’t really work for me, for 2 reasons: not all filters they describe are included in the Stobists collection and the light bulbs that I have do not quite match any of the categories they describe (they are 2700K, which makes them candidates for F1, but they are called soft-white, which makes them F2). Here is how the experiments look like here:

1. Put some gels on the flash
2. Shoot the gray card (using the already established color balance for the gray card), overpowering the existing light
3. Check the RGB histogram to see how far from gray you are (hint: the neutral gray has exactly the same histograms for all RGB components)

The experiment shows that I Roscosun 85 (3/4 CTO) +1/2 CTO bring you very close to the desired color, except for the missing green. This can be confirmed by the formula in the brochure too. Except that there is no enough green. So throw in a PlugGreen gel on top of the CTOs and the match is almost perfect.

The next step is to switch the color balance to the color of the flash (with the gels). Just follow the first procedure above. This is to ensure that the color of the objects hit by your flash (the foreground) is perfect and it is the background that is a bit off, instead of the other way around. Obviously you want to include some portion of the ambient light into the mix for the white balance too, since that is going to be the light on the main subject. If you are planning to bounce the flash from the wall or the ceiling you better include that into the mix too.

After all that you can shoot a picture that includes you available light bulbs into the mix. If everything went as designed they should show perfectly while (for the part of them that is not overexposed of course.)

Are we done? Not quite. There are 3 caveats that I know of so far:

1. Make sure there is no daylight coming through the window. Any daylight that comes into the picture looks blue, which is ugly.
2. The combination of filters described above reduces the amount of light by 2 stops. If you forgot to buy the most powerful flash on the market and a F/2.8 lens you may have to use very high ISO in order to get proper exposure.
3. I normally shoot using my main flash off camera, triggered optically by the built-in flash. The built-in flash is color balanced as a daylight (see #1 above), so unless you attach 3 gels to it this technique is out of the question. If you just tape the 3 gels onto the built-in flash you run into a problem if you want to close it, but I don’t see many other alternatives.

Isn’t it much easier to just overpower the existing light with your bounce flash (preferably off camera) and forget about the color balancing altogether? That depends on 2 factors:

1. Is there anything interesting in the available light that you want to preserve in the picture
2. Can you overpower the available light (in a square room with white walls and white ceiling that is relatively easy, but there will be a lot of cases where the task is much more difficult, e.g. very high ceiling or colored walls/ceiling.)

At least for me the choice is clear – if I can, I will overpower the available light. The balancing is just too much work for a very minimal payoff.

If your budget for toys is tight here is a much cheaper solution, which a friend recommended.

We started from the reviews on Photozone. They have both lenses compared on both EOS 350D and on EOS 50D. In theory the more expensive lens is better in any aspect – it has better (may be 15% difference on 50D) sharpness, almost twice lower vignetting and almost twice lower color aberration.

The theory is a good thing, but $1000 price difference is more than the total price of our photo equipment. So there are 2 questions:

• is the difference visible (can we point to it on a picture)
• assuming the answer to the first question is Yes, are we prepared to pay extra $1000 for it

Enough discussion, here are the pictures. I took multiple shots from the same position. The camera settings are all manual except focus, so the only difference between those pictures is the lens. However I did take the picture is RAW + JPEG mode, so I can show you both the JPEG produced by the camera (EOS XSi/450D) and the picture after certain adjustments made to the RAW file in DPP.

Picture 1, Camera JPEG, 55-250:

Picture 2, Camera JPEG, 70-200:

Picture 3: There is third image into play, made with 55-250, but with different aperture – F/9 instead of F/5.6 as on the first 2 pictures.

Here are the same picture after minor adjustments in DPP. They all have the same settings for sharpness (unsharp mask 3-7-4 and another 100 on the RGB tab)

Picture 1, from RAW, after Lens corrections (peripheral illumination, color aberration (red to the max), vignetting and color blur):

Picture 2, from RAW, nothing changed but sharpness (no visible change from the lens corrections, seems that the lens is perfect as it is, although 70-200 F2.8 IS has better sharpness, half the vignetting and half the color aberrations, according to Photozone)

Picture 3, from RAW, the same adjustments as for picture 1:

I know, spotting the differences between those pictures at 1MP resolution is tough, so here are a few 100% crops, that show the problem areas. The pictures above are just to give you the general idea, although the trained eye can spot one of the issue even on them.

There are 2 main things that we saw:

1. Focus / DOF difference
2. Color aberrations on Picture 1

The focus was supposed to be in the center of the frame. I specifically picked choose the center point and pointed at the center of the circular ornament below the window. That one looks pretty much the same with both lenses:

Picture 1, camera JPEG:

Picture 2, camera JPEG:

Somehow on Picture 1 you see the tree branch in front of the window (the distance is about 3 meters) is clearly out of DOF, while the tree behind (distance of may be 10 meters behind the focus point) is clearly on focus. The DOF calculator shows that at F/5.6, 135 mm lens and 25 meters focus distance (approximately) the DOF is 7.82m, out of which 2.61m before and 5.21m behind the focus point. So the front tree should be just outside of the DOF while the back tree should be definitely outside of DOF. That is much closer to what we see on picture 2. Seems that 55-200 did not focus where it was supposed to. Picture 3 shows that by increasing the DOF (at F/9) 55-200 can produce a reasonable image for the front tree too.

Picture 1, camera JPEG:

Picture 2, camera JPEG:

Picture 3, camera JPEG:

The pictures produced from the RAW still show the same results – the sharpness improves, but Picture 2 is still visibly better than the others:

Picture 1, from RAW:

Picture 2, from RAW:

Picture 3, from RAW:

The back tree is visible on the crops below.

The experiment with the focus chart (see the links section) shows both lenses have a perfectly centered focus from about 1.8 meters. I don’t have an explanation for the result, please contribute if you do.

The color aberrations are usually hard to spot, so we had to actually look for them. Here they are:

Picture 1, camera JPEG:

Here is how the world looks like without them

Picture 2, camera JPEG:

The color aberrations are something that every review says it is easily correctable in software. Well, it was not that easy. The last time I tried clicking the check box in DPP was sufficient, but this time it did not produce any visible impact. I had to actually play with the Red slider below it and to enable the correction of the color blur in order to achieve this:

Picture 1, RAW corrected:

It does seem better that the JPEG from the camera, but it is not comparable to the Picture 2 even remotely.

So, the answer to the first of the 2 questions is quite unexpected: Yes, we can see the difference between the 2 pictures, but:

• on 1280×1024 pixel 19″ monitor we cannot see the color aberration, but we clearly see a DOF problem in the cheap lens
• at 100% zoom the color aberration is more than visible AND is not as easily correctable in software as every review claims.

Before I answer the second question let’s look at another example.

55-250, in-camera JPEG:

The same image, from RAW after sharpening and correcting color aberration and color blur:

And the same picture taken with 70-200 F4 IS, in-camera JPEG

The last 2 images are pretty close, aren’t they? But so what?

Here is the trick. If in the next 10 (or 50) years you are going to take 10,000 pictures with this lens and you would like them to look like the last one it is clearly worth it paying $1000 more today and save all the processing time by shooting in JPEG. If you are going to process these pictures from RAW anyway for other reasons adding the lens correction is not that difficult and could be done on multiple pictures at once (the extra processing time for the lens correction is insignificant, even when multiplied by 10,000). If you never look at your pictures at 100% then you don’t have to worry about the color aberrations. After all this, let’s not forget about the DOF issue discussed earlier.

UPDATE

One more experiment that I missed yesterday: focus/sharpness test. According to the reviews quote at the beginning of the post 70-200 is about 15% sharper than 55-250. I tend to agree with that based on the image up to now.

How do you explain the following shot then:

70-200, 100% center crop, in-camera JPEG:

55-250, 100% center cut, in-camera JPEG:

Both the sign and the license plate on the car behind are sharper on the second shot. Why?

• my Helios 58mm F/2 lens with the coating was about 2 times the price of the same lens without the coating. The same is true for the CPL (circular polarizer filter) today.
• “Don’t put your fingers on the front lens” is still engraved in my mind and the main reason for that is that cleaning the fingerprint may damage the coating

But I never knew what the obsession with the coating is all about. The 0nly reas0n for it that I knew is that it eliminates or at least improves the flare. But you don’t really have to shoot against the sun and even if you have to do it you can usually eliminate the flare by blocking the sun with your left hand (the lens hoods where nowhere to be found at that time.)

Just a couple days ago a friend of mine gave me this article: Colour rendering in binoculars and lenses. I was not really interested in the subject but had some spare time. It is a rather long and boring article, but just around the middle of it there was this distinct “a-aha” moment. In a couple of words the problem is the loss of light (about 5%) due to reflection when the light bean travels between air and glass or glass to air. The first problem with that is that the effect is different depending of the light color and the second one is that problem quickly compounds when the light goes through 10 to 20 glass elements that the typical lens contains today. So, the coating reduces the loss to about 2% and the multilayer coating can reduce it to 0.2%. Well, this IS a big deal.

I did not have many options, so I did 2 things:

1. While in the church I made sure the flash is bounced from the part of the wall that is not covered with paintings yet. That produced a reasonably consistent color tint to all my pictures.

2. Based on a picture taken outside I picked the whitest spot on the guest cloths and used it with the color picker to achieve custom white balance in DPP. Then copied the recipe to the rest of the pictures.

I am quite pleased with the result and the fact the light from the chandeliers is yellowish looks the way I have seen it in a lot of postcards.

Can it be done any better?

Technically the answer is yes, but in practice no one will allow you to put Rosco Sun 85 filter on all church windows in order to bring the sun light to the color of lights inside. Unless you can schedule the ceremony at night (to eliminate the sunlight coming from outside) or turn off all the lights in the church you are basically out of luck.

If you are shooting during the night and there is no sunlight to worry about the Rosco gels are your friend. You can used them to bring the color of your flash to the color of the available light (assuming all light sources in the room are the same color) and then perform a custom white balance in DPP. That will make your pictures look like if they were shot with daylight and completely different from the pictures that your friends are used to see. I will not be surprised if most of them find your pictures to be odd.

To learn more about the color gels and how they are supposed to be used read this wonderful and very technical brochure from Rosco. To get the The STROBIST™ Collection of Rosco gels check the list of resellers here.

Now that I started processing the pictures I decided to see what is the effect of the “Apply Dust Delete data”. Well, there is absolutely no effect. All I get is a message “There was nothing in the image to retouch”. That is reassuring, but I decided to dig a bit dipper into the subject.

The most interesting thing that I found out is that the dust is a problem only at very high F stops. Everyone talking about the subject talks about F/22 and above. My personal experiment shows no visible effect at F/22, but I don’t get the same message, so the software believe that it achieved something. Here are the 2 pictures, I challenge you to find the difference:

Well, I don’t give up so easily. At F/32 the difference is visible (click on them to see them bigger, saving them to a file system so you can easily go back and forth between them works best, although opening them in 2 browser windows should work too):

The experiment shows that I get the “There was nothing in the image to retouch” message at F/11 and below. I have not tried F/16.

Now look at all this from the perspective of DLA. Unless you are doing some serious macro photography and that forces you to work at apertures above F/22 the Dust Delete Data will not help you – you can safely forget about this feature.