Monday, January 24, 2005

Christmas Toys - Part 2

So, after all of the troubles I wrote about yesterday in my quest to get an updated version of my Sigma 500 DG Super EO-ETTL flash, I finally can get to the details of that and the rest of my presents.

The Sigma flash is actually quite a nice unit. For only $200, it holds up quite nicely next to the Canon 550EX flash (which runs about $100 more). It has nice range, supports ETTL (and ETTL-II...as I discussed yesterday), second curtain synch, and a bunch of other stuff. One feature which it supports (which the Canon doesn't, at least as far as I could tell) is that it can serve as an optical slave. That means that you can have the Sigma detached from the camera, and it watches for another flash to go off and then triggers itself. This is nice for getting 2 angle flash coverage. It doesn't seem possible to trigger the Canon 550EX remotely using the Digital Rebel's built in flash. The downside to using the optical slave is that the Digital Rebel does a pre-flash for determining ETTL exposure. That pre-flash is done just BEFORE the exposure begins, and unfortunately it also triggers the Sigma so that it goes off before the exposure. The work around for that (at least for now) is to hold the flash exposure lock button (FEL) on the Digital Rebel. That gives off a pre-flash when you press the button but (as long as you keep the button held down) doesn't do one before opening the shutter for exposure.

One downside I've noticed so far with the Sigma vs. the Canon is that the Canon has auto focus assist for most (all?) of the auto focus point, whereas the Sigma only has center focus point assist.

In addition to receiving the Sigma flash from my wife, she also got me a Sto-fen Omni Bounce for the flash. The omni bounce is just a translucent piece of white plastic that fits over the head of the flash and diffuses the light from the flash (at the expense of a few stops of light). I haven't had the chance yet to experiment with this and see just how worthwhile it is (or if it was just a waste of money). I'm considering getting some other bounce attachments for the flash. Possibly an 80-20 bounce unit. The purpose of such a unit is that you configure the flash head to bounce off of the ceiling, but then you put this attachment on and it lets 80% of the light bounce off the ceiling like it normally would, but it immediately bounces 20% of the light back forward. It gives you some stronger direct lighting while still providing plenty of light to bounce around filling in shadows.

The last gift my wife got me was a set of Pro-Optics extension tubes. These tubes allow you to focus with an ordinary lens at a much closer distance, giving you extra magnification and essentially turning it into a makeshift macro lens. The extension tube set comes with 3 tubes: 13MM, 21mm, and 31mm. You can use the tubes individually, or combine them in any combination for up to 65mm of extension. A lot of people have been raving about the Kenko extension tubes, but these tubes are a much better deal. They are about $40 less expensive, and unlike th Kenko tubes, they actually support EF-S lenses right out of the box, without making any cutting modifications (and voiding the warranty) like the Kenko requires. The only advantage I can see of the Kenko's is that they are 12mm, 20mm, and 36mm for a total of up to 68mm of extension...ever so slightly more the the Pro Optics, but that's so slight it's not worth the extra cost and lack of EF-S support (at least in my book). I've done a little but of work with these so far, but I'm just now getting around to it (since I really needed to have my flash exchange resolved first to use it to work with these things). I'll post some updates and samples later this week.

Lastly, on my Christmas list I also had several other things I didn't get. It seemed everyone would rather buy me video games from the store rather than order stuff online. However, there were some books there that I really wanted, so the week after Christmas I went and bought them as a gift to myself. I picked up "Understanding Exposure" and "Learning to Set Creatively", both by Bryan Peterson, as well as "Night Photography" by Andrew Sanderson. I've already finished the first book and am just now getting started into the second on. I'll post some of my thoughts on those in the coming week or so.
...click here to read more!

Sunday, January 23, 2005

Christmas Toys - Part 1

OK, so it's already a month past Christmas and I'm just NOW getting around to saying what I got? Well, bear with me, because I have a pretty good excuse.

When I created my list, I picked out a number of photography accessories I was interested in, including some books, some cokin filters, extension tubes, and a flash. The flash was the major ticket item on the list. I was interested in the Sigma EF-500 DG Super flash. It is very much similar to the Canon 550 EX flash (a few things better, a few things worse), but $100 cheaper. However, there was only one problem...Sigma released a new version of the flash with the same model # but a different part #.

According to some of the Sigma tech support people and the FAQs on their web site, it seemed that the new version of the flash added support for ETTL-II, which the old version was missing. Being that ETTL-II is supposed to be a function of the camera, that didn't seem to make a whole lot of sense. However, it is worth noting that Sigma builds their Canon equipment by reverse engineering the hardware, not through official licensing. Therefore, it is conceivable that they reverse engineered the spec incorrectly, and that mistake only shows itself once you try to take advantage of newer feature. In fact, this very type of thing has happened several times before, where a Sigma lens which worked with older Canon cameras fails on a newer model and needs to be sent back to Sigma to be rechipped before it will work. So it seemed very possible that this is indeed what happened with this flash.

Now, the Digital Rebel doesn't have support for ETTL-II, but since I anticipate upgrading my camera body at some point, it only seemed to make sense that, given the choice of two flashes otherwise identical in features and price, I should choose the model with ETTL-II support.

So, back to the issue of locating one in stock. Looking around, Amazon was the only one I saw specifically listing the new part number, but they were continually out of stock. Then, after keeping an eye on the flash at several places, I noticed that B&H delisted the flash with the old part numbers and listed a new flash with a different B&H part # and the Sigma part # corresponding to the new model. That seemed like a good sign, so I told my wife that she could now order the flash from B&H if that was what she was going to get me.

So, Christmas arrives, and I open my gifts and find (among other gifts, which I will get to in my next update) the Sigma flash I asked for. The box said "EO-ETTL" on it (which was an indication of the new model...the old one just said "EO"), however the flash on the inside was only labeled EO. It seemed someone had put an old model flash into a new model box. A quick check with Sigma tech support indicated that this was indeed the old flash, and that it would NOT support ETTL-II. Checking with other people on internet message boards, it seems many people had the same problem, and got the same advice from Sigma. One of the higher ups from B&H even chimed in and was also under the impression from his conversations with Sigma that there was a difference. So with that, I contacted B&H to arrange for an exchange of the old version for the new version. I sent back the flash and included a notice indicating how to identify the new model flash and to ensure that I get the correct one this time.

A few weeks later I finally receive my replacement, only to find that once again I received the old flash in the new box. So it was with displeasure that I contacted B&H to arrange to resolve this problem, only they couldn't guarantee that I would get the correct one next time. At that point, I was prepared to simply ask for a refund, however, that's when the story took an interesting turn. B&H contacted Sigma once again, this time speaking to the company president, and Sigma seemed to reverse their stance on the issue, indicating that the differences were purely cosmetic. That seemed mighty shady...like they were trying to cover it up while unloading all the old inventory on unsuspecting customers. However, there was also another detail...several people with the 20D camera (which DOES support ETTL-II) were claiming that they were seeing definite improvements with the old flash on the 20D versus the Digital Rebel, a sign that ETTL-II was working correctly. In addition, one person indicated that they saw absolutely no difference between the old and new versions of the sigma flash when used on the 20D, further evidence that they were the same. However, I imagined that there could still be some differences that might only show up under special conditions. Also, in spite of Sigma's president's statement, their web site was still indicating that the old model flash did NOT support ETTL-II.

At this point, I didn't know what to thing. Then a few days later, Sigma finally updated their FAQ to indicate the flashes were the same, and the only differences were cosmetic. With this public endorsement of the facts, I was ready to take off my tin foil hat, forget the idea that sigma was trying to pull a fast one, and accepted the idea that Sigma had just made a mistake, and the flashes WERE the same in every way that actually mattered to me.

Luckily, I was still waiting on B&H to send me the UPS label to ship my second flash back to them. In light of the new evidence, I decided to cancel the return. So, after getting my flash for Christmas and not being able to use it for nearly a month, I finally was able to take it out of the box and slap it on my camera. Tomorrow, I'll update with a little about the flash as well as my other Christmas presents.


...click here to read more!

Thursday, January 20, 2005

Images Are Fixed....Somewhat

Well, it took a bit of a hack to do it, but I managed to get the images working. And contrary to what I said, my hour or so of scripting was NOT a waste. In fact, it was a nice start. With a lot more hacking, I have a nice generic script working to serve links to any of my pbase images, and it does it automatically, so I don't have to add any files or settings for each new image.

What the script does is take an image ID and size from the URL query string. It then uses those parameters to construct a url to request from pbase. This url returns the pbase page in which the specified image of the specified size is displayed. The script then parses out the image url from the rest of the text and html. Now, at this point, my instinct was to tell the browser to redirect to the pbase image url. That seemed to work fine when viewing images, so I built in a bunch of intelligent url caching and such. Then when I put the links into a web page, I ended up with one problem. When redirecting off a single image or a local html page, the web browser doesn't specify a referrer. However, when viewing an image embedded into a web page from a web site, the referrer DOES get sent to pbase's server. Well, it turns out that some of their servers reject your request when there is a referrer from another site.

So my next attempt was to rewrite the script to have it download the images and cache the entire image locally. When the script does the request, it's sure not to pass along a referrer, so all works fine. With that problem solved, the last issue was what to do about the resized images that were missing. My solution there was to attempt to use the real image of that size from pbase, but when the image was missing, I found the largest available sized version of the image, downloaded, resized it locally, and saved it to a file. This seems to work fine, because I've yet to find any images where the original size was missing. However, I coded the script to handle that situation if it does occur. If the best image available is smaller than the size requested, I don't bother to upsample (what's the point). I just return the smaller image.

So, a little programming trickery and I've now hacked it to work. Now back to...what was this blog supposed to be about...oh yeah, photography!
...click here to read more!

Monday, January 17, 2005

Update on Broken Links

Well, it seems that the problem with the broken links is actually due to some serious problems over at pbase.com. It would seem they had some drive failures and have been having problems recovering. From what I can gather, people have been complaining about this off and on for a few months now. No word from them when they expect to have it fixed.

To resolve the issue with the links, I started working on a script that would take the image request, figure out how to get the image from pbase, and substitute the URL in it's place. Seemed to work fine with my test images, but as soon as I went to try using it to fix a blog entry here, I discovered that it's not just a problem of broken pbase external links. While that is a major component of the problem, it also seems a bunch of the images are missing in the resampled sizes, which are what I post in these blogs. As a result, the hour or so of scripting that I just did appears to have been wasted, as not all the images are there.

Even programming prowess and trickery can't seem to overcome this problem. I might just have to give in and find a new place to host these images. I'm disappointed that pbase can't get this fixed any sooner.
...click here to read more!

Sunday, January 16, 2005

Broken Links

Well, looking back through my blog, I realize now there are a bunch of broken links. I've seen people complaining about linking to pbase lately, but never got around to checking out my stuff. Now I see a lot of it's broken. Now I have to go through and see what I can do about it.

Also, I just had a though...my month+ long absence from posting updates combined with the broken links probably made a few people think this blog was now officially history. Well, it certainly wouldn't have been a record for the fastest site to go stagnant, but I'm not about to throw in the towel here after only 3 months. So I apologize again for the lack of updates. I promise to do better.
...click here to read more!

Getting Back Into Things...Custom Color Profiles

Wow, long time since my last update. Really didn't mean for this to happen. Time to sum up what's been going on.

Since my last post in December, I spent some time working with Wayne from my photography class (which has now ended). He picked up a GretagMacbeth Color Chart, and we did some work with Adobe Photoshop setting up some custom RAW conversion profiles. Normally I do all of my RAW conversion with C1, which doesn't give you the option of setting up custom profiles (or at least the LE version doesn't...not sure about pro), but it does have some built in profiles, including one for the Digital Rebel. I really couldn't imagine having to go back to Photoshop's RAW workflow (very primitive and clunky compared to C1). However, my goal was to learn how custom profiles work, learn how to create them, and then decide if the difference was significant enough to consider using Photoshop for RAW conversion, either full time or (more likely) just to process my most important images.

The first step is to lay out your color chart, get it very evenly lit, and then take your shot. You want to get the chart as straight on as possible (without shadowing it) and to get it exposed as hot as possible (without blowing it out). After you do this, you then load the image in Photoshop's RAW converter, set the exposure, set white balance, then adjust brightness contrast and shadow, and finally the saturation. The goal in the above adjustments is to get certain patches on your exposure of the color chart as close as possible to the reference numbers on the source color chart. Finally, you go through the settings on the calibrate tab, adjusting them all repeatedly until everything is as close as possible. Its a real pain to do it, but when you are done, you should end up with a reasonable close match.

Well, our first time through it, for the life of us, we just couldn't get anything to match up very well. We went strictly by the numbers, and when we finally got things not perfect, but as good as they wanted to get, our colors were horribly wrong. We spent a couple hours working on this without success before finally calling it quits for the day. My hope was that I'd be able to go home and work on my own a little more freely and be able to figure out what went wrong.

Well, it didn't take me long at all to figure out what happened. The reference chart we were using was designed in the PhotoPro RGB color space, and the (R,G,B) values printed on it were the values in that color space. However, the images we were working with were in AdobeRGB color space, and we were trying to get the (R,G,B) values in AdobeRGB to match the (R,G,B) values for PhotoPro. A recipe for failure.

So, feeling stupid enough, I got back on track, created color charts for working in AdobeRGB and sRGB space, and was able to hammer out a reasonable configuration. However, it was a real pain and very nerve racking since you can never get it perfect and changing a setting to make one color a better match ends up making another color a worse match. My next attempt was to download the ACR-Calibrator script written by Thomas Fors. It goes through the process of RAW conversion iteratively changing each parameter a little at a time until it finds the best set of parameters. Seems like a quick and easy script to run. Well, it's certainly easy, but not quick. The process took over 2 hours on my AMD 2800+, as it took nearly 900 RAW conversion to come to a solution. And the results were quite promising, matching much better than my by hand configuration did.

So, how did it end up? Well, the grayscale patches on the chart matched up very well, but the colors were hit an miss. Some were nearly indistinguishable from the source chart, but other were way off. So after all of this, the real question was, how did it hold up to C1? Well, it was a very mixed result. First off, it feels like a really unfair comparison since Photoshop gives you all these option to adjust it precisely, and with C1 you really just have to eyeball things and make a few slight tweaks. However, you have to remember that comparing what kind of results you get with the options available to you is what this whole test was about. So with any unfairness safely out of mind, here is what I felt after coming up with my best approximation in C1.

The grayscale patches were definitely closer in Photoshop, but the color patches were a toss up. Overall, I think Photoshop did a better job matching the brightness of the colors. C1 tended to give brighter results on many, but reducing the brightness in C1 to make those colors match would just make others darker. I don't doubt that with enough tweaking (contrast, curves, etc) I could get C1 looking better (and this would probably give better results for the grayscale patches too). On the other hand, when it came to the tone of the color, I think C1 tended to give better results. The results from Photoshop just seemed to have strange tints to them.

So in the end, I didn't seem to be satisfied with any real improvements in Photoshop. I think I'd be more likely to get a better match in C1 by just adjusting the levels and such. In addition, this custom profile business seems very theoretical. Yes, you can get a profile set up for a shot, but realistically you should have a custom profile for each ISO setting under each different lighting condition. That seems like too much of a pain. I'd probably end up with just a few profiles for different conditions, and then use the one that is closest. I'm not sure that approximating in Photoshop is necessarily any better than approximating in C1. Seems like you'd lose out on the extra flexibility that Photoshop offers.

Finally, when it came down to it, I started thinking about how important the whole idea of precise color matching is anyway. Certainly there are those really art oriented people, who will raise a fuss if the colors aren't exactly as they saw at the time...wanting to be true to the spirit of the original subject and all. Myself, I seem to find myself happy with close approximations. Who's to say that the "true" colors are the best. I can certainly come up with photos that give me satisfying results in C1.

So in the end, it might seem like wasted time, but it was educational, and it was also reassuring to prove to myself that C1 is indeed the right tool for me. It's never bad to go back and reexamine your tools. As you gain experience and knowledge, you may find yourself outgrowing tools that seemed fine before. It felt good to prove to myself that C1 is NOT one of those tools I have outgrown.

So, that was my experience working with custom color profiles. In a few future posts (which I'll really try to get up this week) I'll mention my other year end/new years projects and my Christmas goodies.
...click here to read more!