GX7 Lens Review on DxO

Started Dec 15, 2013 | Discussions thread
Alexis D
Regular MemberPosts: 395
Like?
Re: DxO's Comments on this...Strong AA Filter?
In reply to Ken Strain, Dec 19, 2013

Ken Strain wrote:

Anders W wrote:

Steen Bay wrote:

Anders W wrote:

MarkyM wrote:

Emilie from DxO has commented on their site saying they "guess" it is a strong AA filter:

"Yes, this is something we checked many times before publish it,...
Sharpness measured on GX7 is a bit low, but pretty close to the GX1 results. We guess there is a strong Anti Alias filter on this cameras.u

It seems to me they did something wrong or had a bad unit.

Strong AA filter as an explanation for this kind of discrepancy (worst results for any MFT camera ever tested, including all those with 12 MP sensors)? No way. If that had actually been the case (which is of course extremely unlikely in the first place), it would have been known long ago. Several test sites (inluding SLR Gear/Imaging Resources and optyczne.pl, the Polish counterpart) have looked at the resolution of the sensor without finding anything unusual.

So DxO made a mistake or had a faulty unit. What is really weird is that they went ahead and published the results in spite of the fact that something is clearly wrong.

Checked.. DxO has tested 29 different mFT lenses on 13 different mFT cameras (+ a few older lenses only on older cameras), and with every single of the 29 lenses the GX7 gets a very low score. Makes me wonder... mistakes happen, but 29 times in a row! Could it be that they don't actually test every single lens on every single body, but instead use the test result from one camera to derive/calculate the 'results' for all the other cameras?

No, I wouldn't think they'd dare to go that far (or that it is possible to do so "safely").

If doing it like that, then it's only necessary to test a single 'reference' lens on a new camera like the GX7. Saves a lot of work, but if something goes wrong, then all the derived results will be wrong too.

I initially suspected that they'd run into some kind of shutter-shock problem, perhaps some problem of that kind more or less unique to the particular copy of the camera they tested. On the other hand, one would then expect more problem with vertical than horizontal resolution, and I didn't see anything suggesting that pattern when I had a look. OK, I certainly didn't check very thoroughly but anyway.

It could be something as mundane as a programming error, however. If they entered the wrong parameter/constant when going from their raw observations to the final acutance and P-MPix scores, all the values will of course be off in the same direction.

I am still not sure we can entirely exclude the shutter-shock hypothesis, however. I had a look at the review of the GX7 by optyczne.pl (Lenstip but in Polish) and they apparently ran into a problem with suspiciously low resolution values which forced them to switch to the electronic shutter to get proper values in that department. See here:

http://www.optyczne.pl/239.4-Test_aparatu-Panasonic_Lumix_DMC-GX7_Rozdzielczość.html

Hi Anders,

A few observations and comments (no conclusions yet):

I checked the sigma 60mm at f/4 with the EM5 and GX7 on DXOmark for PMpix and acutance at the image centre at about the sharpest aperture:

  • It is likely that the relation from PMpix to some MTF measure involves a square root - so the linear measures would differ by sqrt(11/6) ~ 1.35+/- 0.02 (error from the rounding of PMpix to one significant figure).
  • Out of curiosity I applied my model for estimating MTF50 from the DXO acutance, and get 945 : 716 cycles/picture height, or a ratio of 1.32.

The GX1 is much closer to the EM5 on DXO (within 10+/-5% in linear measure). So the quote from up-thread that the GX7 is "pretty close to GX1 results" is hard to understand. It looks like an outlier to me.

Although I have only measured the vibration of one GX7 (and no EM5s), I will be a little surprised if it is concluded that there is a general problem with the GX7. There is the possibility that internal resonances of particular mounts (tripod/head/etc.) interact with the internal dynamics of the camera to make things worse.

It is pretty bad news if even half the discrepancy is from the measurement, as that spans the range from a good lens to exceptional one.

Ken

ps., mildly amusing (perhaps) trick question: which has higher "acutance" a Sigma 60mm at f/4 on the EM5 or a Zeiss APO Sonnar 135/2 on a Nikon D4 at f/8? (Apertures were chosen as the ones I would use most often, and of course the cameras both have 16 Mpixels).

Answer: neither - the Sigma/EM5 combination is marginally better in frame centre, but marginally worse at the edges). PMpix scores are 11 and 15, respectively. I think the price and weight of these combinations differs by more than the qu

ality of the result.

At the moment the whole business seems like a bad joke.

Trick question not so tricky. I suspected some answer like that.
I think DxO is confusing real test results with expected results and just publish expected results. Surely big dollar product and big brand products with big marketing pockets are expected to have better test results.

Reply   Reply with quote   Complain
Post (hide subjects)Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark post MMy threads
Color scheme? Blue / Yellow