dpreview Test

[t024484]

I found this test, which included the M8, interesting because coloured resolution charts were used, as well as b&w.
It was the first time I saw this type of test.
I'm not sure, but I fear the shots were taken as Jpeg's, wich is bad for the M8.
Nevertheles, apart from dark blue, the M8 was doing very well against a 10 Mb D60 and a 14 Mb Foveon sensor

Sigma DP1 Review: 20. Compared to...:

Hans

[jimmy pro]

Wow! Well if those are jpegs then obviously you've got processing algorythms as a variable, but damn, the Nikon D60 looks as good as the M8 all the time, and sometimes better. And its not got a Leica lens either. Quick, somebody think up a bunch of great excuses why the M8 is really better, before I'm tempted to sell mine and get a D60

[jaapv]

No excuses, but the test was discussed extensively here, and all Jimmy's excuses were trotted out , to no avail. The end conclusion was, however, that this kind of test, although maybe interesting for sensor technicians, has very little relevance for actual picture taking.

[stunsworth]

I disagree that the D60 looks better. If you look at a lot of those shots the lines in the chart merge into a mush for part of their length - presumably due to the anti-aliasing filter doing its stuff - look at the blue lines on a yellow background as an example.

[t024484]

Originally Posted by jimmy pro

Wow! Well if those are jpegs then obviously you've got processing algorythms as a variable, but damn, the Nikon D60 looks as good as the M8 all the time, and sometimes better. And its not got a Leica lens either. Quick, somebody think up a bunch of great excuses why the M8 is really better, before I'm tempted to sell mine and get a D60

Maybe there is something wrong with my eyes, but apart from being a bit soft, all M8 images were superior.
I either fail to see why this test has no relevance to the real world.
We live in a coloured world, not b&w.
So if you want to test resolution, a coloured resolution test seems the better one to me.
It was the kind of testing that I liked, not the test itself.
I am looking forward to see the same M8 tests in Raw, with top contenders from N & C.

Hans

[jimmy pro]

Hey, thanks for the emergency shot of heavy-duty denial! Just what I needed. I'm definately keeping the M8.

[Philinflash]

Excuse me, but are not lenses a factor in a test like this?

[t024484]

Originally Posted by Philinflash

Excuse me, but are not lenses a factor in a test like this?

What would be your guess ?

[stunsworth]

No denial here Jimmy, if you think the D60 looks better on the two blue/yellow sets then fair enough. But as I say, look at the clarity in the lines in the Nikon shot and I think you'll agree that the Leica is better.

[jimmy pro]

I see your point on the blue backround w/yellow printing, but not on any others. And even on that one, if you stop to consider the M8 is useing a Leica lens and the D60 is not. I kinda wish they said what Nikon lens was used. If it was the kit zoom then its even more likely that CA is hurting the D60 shots and with better lenses it would leave the M8 in its dust. And I'm familiar with how in photography, at the upper end a little improvement costs a lot of doe, but still, for how much more the M8 costs then a D60 it should of wasted it, but it didn't.

Bottom-line is this, the ergonomic and/or rangefinder-vs-slr reasons for preferring the M8 are the only one's you don't need a very active imagination to make.

[Philinflash]

Originally Posted by t024484

What would be your guess ?

My guess would be that these tests are inconclusive crap.

[SJP]

Short, succinct!

I like it !!!

[ho_co]

Originally Posted by Philinflash

... these tests are inconclusive crap.

Clear and well said.

Was the M8 equipped with UV/IR filter? If not, that would likely explain the variation in color results.

The testers clearly don't have a good handle on the way the M8 is constructed either:

The Leica M8 which has a very light low-pass filter ...

The test results are clear: The M8 sucks.

Does your M8 suck? Mine doesn't.

Maybe the test sucks?

[AlanG]

Originally Posted by jimmy pro

I see your point on the blue backround w/yellow printing, but not on any others. And even on that one, if you stop to consider the M8 is useing a Leica lens and the D60 is not. I kinda wish they said what Nikon lens was used. If it was the kit zoom then its even more likely that CA is hurting the D60 shots and with better lenses it would leave the M8 in its dust...

From DPReview:

Here are the lenses they say they used on the studio scene: (I presume they used the same lenses on the resolution tests.) The rest of the points about the IR filter, lights used, testing procedures, IR, have all been covered in another thread.

"For direct comparisons we always use sharp prime lenses stopped down, typically to F9 for 35 mm lenses. Here we have used the Nikkor 50 mm F1.8 (it's sharper than the F1.4 at F9) on the D60 and D40, the Canon EF 50 mm F1.4 on the EOS 450D and the Summicron-M 50 mm F2.0 on the Leica M8."

[ho_co]

Thanks, Alan. Do they say whether they used any kind of UV/IR filter on the M8?

IMHO, dpreview is comparing apples and oranges and selling them as apricots.

Why do they choose f/9? And how the heck do you set f/9 on a Summicron?

That aperture is already smaller than most of these lenses' optimum.

At least the test gets them a few extra page visits.

[Daniel Ortego]

Damn, I’m glad I didn’t weigh in on this thread.

[AlanG]

Originally Posted by ho_co

Thanks, Alan. Do they say whether they used any kind of UV/IR filter on the M8?

IMHO, dpreview is comparing apples and oranges and selling them as apricots.

Why do they choose f/9? And how the heck do you set f/9 on a Summicron?

That aperture is already smaller than most of these lenses' optimum.

OK, I have nothing to do with DPReview and I'm not criticizing or defending them. But I did read the other thread and additional info at the DPReview site.

They did not us an IR filter, thus there was no extra piece of glass to affect the optical path on the Leica. The fluorescent photo lights they use do not put out IR. Besides, even if IR could affect the image focus, the red channel would show blurring of the black and white test target. But it didn't.

I think the only thing you can question about the red/blue blurriness is that they used in-camera jpegs. It could be that raw conversions wouldn't show this result. The red/blue color blurring is happening somewhere in the Leica's image processing chain not in its light gathering.

They use those lenses at f9 because they are trying to compare the cameras and not the lenses. So they look for the best normal primes by each manufacturer and use them at the aperture that typically gives the best results. If f9 is not the lens' optimum aperture, it is unlikely it is very far from it. It shouldn't matter too much - note that the Canon 70-200 out resolved the 5D sensor at various points. I'm almost sure the Summicron 50 f2 can out resolve the M8's sensor at f9. If not, there won't be much reason for Leica to come out with a camera that is higher than 10 megapixels other than to go full frame. These are similar lenses that are being used within their typical photographic ranges. (Whereas some lenses might be optimized for closer distances and others for farther distance, flat field, etc.) You can't test for everything...

DPReview recently started doing lens tests and they are the most comprehensive lens tests I have ever read. I think their camera tests are quite thorough too. But I do think this M8 color test should be repeated with raw files just to pin down the problem.

[diogenis]

M8 only fails in the dark blue/red test, all the other color tests show clear, artifact free and a bit softer image than all the rest... so what?

So, this test shows some superiority in this particular sensor technology over the others. It is test that's designed to exploit particularly all the goods a Foveon sensor gives without the minuses. It is like some guys from Sigma saying that "hey we have designed a super duper sensor that can take the sharpest pics of a supernova star exploding" well... good for them, but I don't buy it. There are many more factors in sensor technology, and I prefer letting other engineers choose the right one for their cameras.

These tests also show that -I believe they did used jpegs for this demosaic procedure to kick in- there might be a problem to m8's jpeg routines, Leica's s/w engineers might need to take a look into. For those that shoot in jpeg that is (if there is anyone that has a Leica and shoots in jpeg )

[vanhulsenbeek]

Originally Posted by diogenis

These tests also show that -I believe they did used jpegs for this demosaic procedure to kick in- there might be a problem to m8's jpeg routines, Leica's s/w engineers might need to take a look into. For those that shoot in jpeg that is (if there is anyone that has a Leica and shoots in jpeg )

The use of jpegs was confirmed in another thread on this subject. We know that Leica's s/w engineers might need to take a look into the M8 jpeg, but the fact that the test used jpeg's makes it pretty unreliable IMHO anyway

[jaapv]

Now if this test is indeed one that measures two parameters, eg. jpeg processing and sensor response, without defining the relationship between the two, it is utterly invalid. If one wants to compare sensor output between cameras one needs that output to be as uncorrupted as possible. That means RAW data. I suspect that was too much work... I'm surprised that DpReview produces such a sloppy test.