« OT: Forget Not | Main | New Rolleiflex MiniDigi »

Saturday, 15 March 2008


LRViewer is a pretty nifty little application. Basically you can take a Lightroom catalog that has had all the previews rendered, put the catalog on a USB thumbdrive or CD, and then read the catalog on any PC or Mac, without installing Lightroom on that system. If you need a photo from the catalog, just extract it from the previews.
Makes for a MUCH better photo viewing experience than working with an HTML gallery.

DPreview's review of the D300 was like an announcement of the second coming!
(Maybe we have arrived?)

I'd buy one myself if it had image stabilization in the body.

Mike J.

Personally I wouldn't get hung up on in body stabilisation. Firstly there is a limit to the number of lenses one uses and having to purchase them with VR for me wouldn't be a deal breaker if I felt I had to have it. Secondly the higher usable ISO supplied by the D300 compensates for the extra stop or two image stabilisation offers, no ?

Yes it would be nice to have both stabilisation and better ISO performance but c'est la vie.
O/K I'm biased. I switched to a D300 from a Minolta D7. I have VR in an 18-200 lens but not in the Sigma 10-20 or the Nikon 28mm f2.8. At these wider focal lengths I don't feel I miss it too much.

At the risk of hijacking the conversation...

What caught my attention in Thom's review was the mention of AF Fine Tune. When Big Mike and I discussed focusing problems under this topic:


neither of us made mention of a second source of phase detection focus errors, namely the fact that the system predicts where the lens should be focused and if the result is "within tolerance" it doesn't make any further adjustment. (I didn't mention it because I was unaware of it!) When it comes to lens sharpness and focusing, my experience has been that "within tolerance" equates to "unsharp." Being able to fix that to my finicky standards would be a very important factor in choosing this camera.

This bit of information also opens up a whole new can of worms. Focus errors due to the second factor don't reflect quality control and build tolerances in the cameras but in the lenses, a situation we haven't dealt with before. And that has me thinking...

So-called "kit lenses" are often reported as having substantially poorer image quality than higher-priced lenses. That could very well be true. But it's also possible that the problem lies in the quality control for such lenses, which may be lower to keep cost down. I wouldn't be at all surprised if there isn't more unit-to-unit variation, and that would make predictive focus less likely to be spot-on and more likely to merely be "within tolerance."

It would be very interesting to test some of those lenses out on a camera that allows for focus fine tuning. We might be surprised at how good some of those lenses are... or not.

Just a thought. I'm not equipped to do such testing, but if someone else wants to jump in I would sure be interested in what they find out.

~ pax \ Ctein
[ please excuse any word salad. MacSpeech in training! ]
-- Ctein's online Gallery http://ctein.com
-- Digital restorations http://photorepair.com

"the higher usable ISO [...] compensates for the extra stop or two image stabilisation offers, no ?"

In body shake reduction is stated to obtain a 2 to 4 stops benefit. At least with the Pentax K20D. Which also allows AF adjustments for up to 20 individual lenses, or globally.

It might not have the same high ISO quality of the D300 (at ISO 3200 and 6400). But real world use reviews and examples have shown ISO 1600 to have a 2 stop improvement over the K10D, and even ISO 3200 (and 6400) being usable -- if you know where it's limits are. And it's image quality is receiving high praises, such as the one in the latest Chasseur d'Image print edition.

So high ISO usability and in-body shake reduction allows for even better hand held low light photography -- if compared to high ISO usability alone.


>> Being able to fix that to my finicky standards would be a very important factor in choosing this camera. <<

As best I can tell, the AF fine tune feature addresses only accuracy, allowing the user to dial in a fixed bias on a lens by lens (or focal length by focal length, for a zoom) basis. It can't do anything about precision.

AF precision has been an issue for as long as such systems have been around. I remember discussions on the old CompuServe photo forum almost twenty years ago, when the original EOS-1, with its innovative cross-sensor, was released along with the ultrafast 50mm f/1 and 85mm f/1.2 EF lenses. It's not clear to me whether even today's EOS-1 DSLRs, evaluated as a system with the lenses mounted, have the combination of accuracy and precision required to reliably deliver optimal performance from such lenses at full aperture.

Dear Oren,

That's a good question. I know very little about how the systems work, so I don't know if the biggest source of error is in the accuracy or the precision.

(For the lay folk out there, precision is how finely you can measure something; accuracy is whether that measurement is right. Good examples would be cheap darkroom thermometers, which you could read to a fraction of a degree (precision) but were rarely accurate to even a whole degree, or the Leica M-6 rangefinder, whose long baseline allows for extremely precise focusing, but the focus isn't anywhere near that accurate.)

Film camera autofocus systems started with very few zones of focus, it's true. Autofocus point and shoot cameras often had only a few dozen or less. They quickly evolved to the point where they had several hundred distinct zones of focus, enough to (barely) support even a 50 mm f/1 lens, had such a P&S camera existed. I have no idea where the DSLR's are today. Either accuracy or precision could be the limiting factor.

Seems to me one could test this in the following fashion: set up the camera on a tripod and position a test target on a table alongside a ruler 30 or 40 focal lengths away, say 1.5-2m for a 50 mm lens. Center the autofocus window on the test target and make a series of exposures with the lens wide open, moving the target 2-4 mm between exposures.

A few dozen such exposures would suffice. You should be able to see when the camera jumps focus by looking at the tick marks on the ruler, and that will tell you the precision of the system. Within each group of exposures in a zone, you can look at the target and see how good the accuracy is ( whether the sharpest photograph the one that is at or near the center of the zone distance range) and how bad the errors are at the extremes of the zone.

I don't have one of these cameras so I can't run this test. But I do have little reflection resolution targets that run from one line pair per millimeter up to around 100 line pair per millimeter in 26% jumps, suitable for doing this kind of test. If someone reading this wants to PROMISE to run this test and will do it rigorously, I'll send them a target if they send me their address. I don't have a lot of these targets, so this is not a blanket offer to send them to anybody who asks (or even anyone at all). Send me your request and I will decide if you're worthy [ smile ].

~ pax \ Ctein
[ please excuse any word salad. MacSpeech in training! ]
-- Ctein's Online Gallery http://ctein.com
-- Digital Restorations http://photorepair.com


i have no intention to do these tests, i just want to say that it is too cool that such issues are discussed here. These are real limiting factors when it gets to ... sharpness , resolution etc.

I often noticed that the camera doesn't "refocus" after i change my position only slightly. I think this limited precision. On the other hand, I couldn't tell a difference then, even with a split-image screen.

The problem is that with high resolution sensors you notice any even-so-small focus errors.

However, nice to read about it...

best always

I also have some doubts that the in-body AF fine-tuning is a cure-all. It seems to me that a fairly large variation of AF results can be seen depending on whether the lens begins its focus from infinity, min focus, or just slightly one way or the other. It would also not be a surprise to me if some some lenses may require different adjustments across differing focal lengths and subject distance.

FWIW I was told by an industry guy in the 1980s that they could make AF systems so sensitive that they would change focus with the photographer's breathing, but that the makers tended to de-tune the systems for two reasons--to accommodate manufacturing slop and to reduce "hunting." He told me that they (he asked at the time that I not associate his comments with his company, so I'll continue to honor that) had made prototypes that essentially hunted constantly, because the photographer and/or the subject just could not be made to stay still enough for it not to hunt. But the greater the potential accuracy, the greater the precision had to be, and manufacturing precision is hard. Interestingly, that company had made and briefly marketed a manual-focus camera with electronic focus confirmation, and the focus confirmation was said to be considerably more precise than the actual AF systems the company eventually decided on. His point was that AF inaccuracy was a design decision, not a technology limitation.

Consider this hearsay, as I'm just going on memory.

Mike J.

I did some informal testing of such focus precision for my Canon 17-55 f/2.8, along ther lines of Ctein's procedure. I found the focus confirmation to be more or less within Canon's stated tolerance (1/3 of depth of focus), which still left a whole lot of variation.

I also did some calculations around the Canon mkII AF "issue". they haven't changed the tolerance spec and at 10fps the subject has to move astonishingly fast to get shot-to-shot movement outside of the 1/3 DoF. No wonder sequences don't generate high hit rates.

Even with my older DSLR, I'm finding I need to use manual focus for critical applications. I think it comes down to the fact that pixel-peeping now means we are becoming much more demanding of AF: the criticality criteria have tightened significantly but the AF tolerances haven't.

The comments to this entry are closed.