Photodo in new wrapping

[jani]

I completely missed this, and I can't recall seeing anyone else posting about it here, so:

Photodo is back as of July (announced eight days ago), but the site has new owners/maintainers.

I like the new site design better than the old one, and it seems like it's on a new and faster server.

The forums is a nice addition.

I just hope that they'll be able to deliver lens tests that are up to the high standards Lars Kjellberg had.

(I'm posting this here because Photodo still is of mostly a technical interest.)

[Ray]

Thanks for pointing that out, Jani. I guess I'm a bit disappointed that Photodo's new tests are 'system' tests rather than lens tests. Perhaps we need some discussion on what those graphs really mean. They seem to be indicating how many pixels it takes to produce one line pair with an MTF response of 50%. For example, the Canon 85/1.2 at around f5.6, in the centre, used with a 20D, produces 0.35 of a line pair (cycle) at MTF 50. In other words, it takes almost 3 pixel to describe one line pair at 50% MTF.

Is that how you read it?

[bjanes]

Thanks for pointing that out, Jani. I guess I'm a bit disappointed that Photodo's new tests are 'system' tests rather than lens tests. Perhaps we need some discussion on what those graphs really mean. They seem to be indicating how many pixels it takes to produce one line pair with an MTF response of 50%. For example, the Canon 85/1.2 at around f5.6, in the centre, used with a 20D, produces 0.35 of a line pair (cycle) at MTF 50. In other words, it takes almost 3 pixel to describe one line pair at 50% MTF.

Is that how you read it?
[{POST_SNAPBACK}][/a]

Ray,

0.35 cycles/pixel is actually a pretty good result, since the Nyquist limit is 0.5 cycles/pixel and Bayer sensors typically yield 60-70% of Nyquist due to demosaicing and other errors.

IMHO, for lens testing one should use the highest resolution sensor and report the result in cycles/mm, but you are still reporting system results. I use Imitest and think it is an excellent program, but unfortunately oversharpening and a weak anti-alaising filter can result in MTFs above Nyquist with Imitest. In this case, one is measuring false detail due to aliasing, essentially garbage, not image detail.

[a href=\"http://forums.dpreview.com/forums/read.asp?forum=1021&message=16693479]http://forums.dpreview.com/forums/read.asp...essage=16693479[/url]

[Ray]

0.35 cycles/pixel is actually a pretty good result, since the Nyquist limit is 0.5 cycles/pixel and Bayer sensors typically yield 60-70% of Nyquist due to demosaicing and other errors.

[a href=\"index.php?act=findpost&pid=72404\"][{POST_SNAPBACK}][/a]

Bill,
I gathered it was a good result by comparing it with the results from other lenses.

BJL, would probably be interested to see that the Canon 85/1.2 (with 20D) displays a greater resolution at a greater range of stops, from around f4 to f8, than the Zuiko 50/2 (with E-500) which peaks at f4 (at 0.3 cycles/pixel) and begins a downward slide at f5.6.

The Sigma 100-300/4 has a surprising result. At 100mm it peaks at a resolution of 0.3 c/p, same as the Zuiko 50/2 except at the much smaller aperture of f13. That's useful to know because on either side of the peak, performance seems to slump dramatically, at f5.6 being only 0.15 c/p and even lower than that at full aperture.

What puzzles me here is, how can a zoom lens have the same MTF response at the same resolution, at an at least slightly diffraction limited stop of f13, as a fine Zuiko prime which peaks at f4, both with 8mp cameras?

How reliable are these results?

[LeifG]

I completely missed this, and I can't recall seeing anyone else posting about it here, so:

Photodo is back as of July (announced eight days ago), but the site has new owners/maintainers.

I like the new site design better than the old one, and it seems like it's on a new and faster server.

The forums is a nice addition.

I just hope that they'll be able to deliver lens tests that are up to the high standards Lars Kjellberg had.

(I'm posting this here because Photodo still is of mostly a technical interest.)
[a href=\"index.php?act=findpost&pid=72382\"][{POST_SNAPBACK}][/a]

At least some of the reviews have already been published on the www.ePhotoZine.co.uk site. As far as I know they use Imatest software. I prefer Klaus's tests on PhotoZone.

Leif

[Ray]

At least some of the reviews have already been published on the www.ePhotoZine.co.uk site. As far as I know they use Imatest software. I prefer Klaus's tests on PhotoZone.

[a href=\"index.php?act=findpost&pid=72540\"][{POST_SNAPBACK}][/a]

Photodo also uses Imatest. They seem to be providing the same resolution test at MTF 50% except the Photozone results are shown line widths per picture height whereas Photodo shows fractions of a line pair (cycle) per pixel.

More significantly, Photodo gives us results all the way down to f32. There are many lenses that perform sufficiently well for landscape shots, beyond f11. There are even some 35mm lenses that are sharpest at f16.

But here's the problem I have with the Photodo result for the Sigma 100-300/F4 at 100mm. The lens shows a remarkable consistency between centre and edge performance which is almost equally good (or bad) across the whole range of f stops, and that's a good result in itself. But even more remarkable is the (apparent) fact that, not only does this lens peak at f11-f16, it is actually equal in contrast and resolution (at these apertures) to the Zuiko 50/2 at its best (at f4).

This seems to me to be either a major blunder on the part of Photodo or a lens manufacturing anomaly. The advantage that an organization has over an individual when testing lenses, is that the organization, if it has a good reputation, can request another copy of the lens from a different batch when results are suspisciously too good or too bad.

If these Photodo results for the Sigma at 100mm are not a mistake, then they really do shatter the myth that diffraction is the main obstacle to resolution beyond f8.

I've included screen copies below of the charts from both Photozone and Photodo so you can see graphically what i'm talking about.

[attachment=869:attachment]                      [attachment=870:attachment]

[Slough]

Photodo also uses Imatest. They seem to be providing the same resolution test at MTF 50% except the Photozone results are shown line widths per picture height whereas Photodo shows fractions of a line pair (cycle) per pixel.

I know Klaus also uses Imatest.  

I prefer Klaus's tests because his results are presented in a more detailed and organised way, including plots for vignetting and distortion. He also seems to have a good feel for testing, and will comment if he sees a sample that he thinks has optical errors that are not a function of the design e.g. decentring.

One example of why I have a bit less confidence in ePhotozine tests is the review of the Nikon 200mm F4 AFD lens. The general concensus seems to be that this is one of Nikon's star performers and I bought one which conforms to that description. I suspect that their sample has aligment/centring problems (having examined the example pictures) but they did not pick them up. Photodo describe this lens as "an adequate performer, but again by modern standards, not outstanding". Well I am not sure what they mean by 'modern standards' as Nikon micros are very highly respected, and most are not recent designs. Some of the designs from 20 years ago perform as well as anything designed today.  In fact some older designs are much better than more recent ones. They also say "Due to the long focal length (for a macro), focussing is critical as very little depth-of-field is achievable" and in the cons at the end they say "Narrow depth-of-field". This is incorrect as DOF is independent of focal length as it depends on image magnification and F number. Okay, I know that these things are not such a big deal, but they do make me feel a little uneasy.

I am sure the ePhotozone tests are very useful, and the tester has used the lenses in the field so I am sure that the reviews are well worth reading if you are in the market for a given lens. But I have more confidence in Klaus's results.

In case you think me a bit sad, I was recently searching for a ~200mm macro lens, hence my scrutinising online reviews. Sad but how else do I assess candidate lenses in lieu of having samples to test myself?

Leif

[Ray]

In case you think me a bit sad, I was recently searching for a ~200mm macro lens, hence my scrutinising online reviews. Sad but how else do I assess candidate lenses in lieu of having samples to test myself?

[a href=\"index.php?act=findpost&pid=72652\"][{POST_SNAPBACK}][/a]

I don't think that's sad. What's sad is the lack of good 'quality control' that often does not inspire confidence in the buyer so that he/she feels compelled to do his/her own testing before buying.

Even a simple chart that accompanied the lens, like the one produced by Photodo, would be useful. I wonder how much it would add to the cost of a lens if such tests were streamlined.

[Slough]

I don't think that's sad. What's sad is the lack of good 'quality control' that often does not inspire confidence in the buyer so that he/she feels compelled to do his/her own testing before buying.

Even a simple chart that accompanied the lens, like the one produced by Photodo, would be useful. I wonder how much it would add to the cost of a lens if such tests were streamlined.
[a href=\"index.php?act=findpost&pid=72669\"][{POST_SNAPBACK}][/a]

I have bought 4 Nikon lenses in the space of 1 year.

24-85AFS: A complete lemon, not suitable for larger than 6x4. Oddly enough subsequent frames can vary, with one very soft and one so-so, despite using a heavy duty support, mirror lock up and a cable release.
12-24 AFS: Superb apart from the wide end (12-14mm) where there is significant edge smearing (not seen in online reviews). At 12mm the overall quality is so-so. Sample variation?
200mm micro: Superb. (Exactly as per most reviews.)
28-105 AFD: Very nice. (Exactly as per the reviews.)

So out of a sample of four, I have had one dog, one below par, and two nice ones. At least one person has talked of trying 4 samples of the 12-24 before getting a good one.

I don't think manufacturers would supply test results for each lens as it would be an admission that sample variation exists, and people would open the boxes to find the best, leaving the others on the shelves.

I know that sample variation existed 20 years ago too as I knew a pro-photographer who would test a boxful of lenses and choose the best.

[Ray]

So out of a sample of four, I have had one dog, one below par, and two nice ones. At least one person has talked of trying 4 samples of the 12-24 before getting a good one.

This is really a ridiculous state of affairs. I'm interested in the Sigma 12-24 for the Canon but have been putting off buying it because I know I'm probably going to spend an afternoon testing a few in the store and I have to find a store that carries a few copies and will let me cherry pick.

I don't think manufacturers would supply test results for each lens as it would be an admission that sample variation exists, and people would open the boxes to find the best, leaving the others on the shelves.

I agree it wouldn't be practicable to provide a test chart and charge the same for each lens. It would have to be a 'value adding' process. You get what you pay for. You buy with confidence and you save all the stuffing around to ensure you get a good copy. If you are not fussy about resolution and MTF performance, you buy Grade D at a really low price. If you want the best, you pay a premium. Is it not always so?