Discussion in 'This or That?' started by JoFT, Jan 21, 2015.
Does anyone knows why the µ43 is performing much worse than APS-C in the scores of DXO??
Well, the quick answer is that all other things being held equal (sensor tech, resolution, etc.) the bigger the sensor, the better your scores/performance is going to be, and APS-C is considerably larger.
Was there a particular comparison you were wondering about?
I am working on a comparison right now. The astonishing is that especially the lenses in combination with the sensor are delivering an impression which is fare away from the reality... The impression you can get looking at the DXO scores is that a µ43 image has
• 10MP out of 16MP: 63%,
• Canon 14/20MP @ APS-C71%
• and the Fullformat 17/22MP...77%
The basis is the following setup
Full Format: Canon EOS 5DMkIII with the Canon EF 1.4/50mm USM
APS-C: Canon EOS 7DMkII with the EF 2.0/35mm IS/USM
µ43: Panasonic Lumix GM1 with the Leica Summilux 1.4/25mm
To measure resolution with lenses, does DXO use RAW or jpg? I thought I read they use jpg, but I'm not entirely sure. At any rate, if true, then the jpg engine could impact the outcome (and therefore could be different between cameras, even within the same format and MP resolution)
AA filter will also impact.
For instance, when I compare the DXO marks for the same lens, but on a GX7 vs. EM10, the EM10 scores better, but I rather think this is down to the jpg engine???
edit: Correct in post #8 below https://www.mu-43.com/showthread.php?t=72799&p=732723#post732723 though no source was cited.
And where do you think this deviates from reality?
The thing you have to keep in mind when you're comparing lenses on DXO is that you're really comparing camera sensor and lens combinations, so sticking that same EF 1.4/50mm USM on a D800E would result in considerably higher scores.
EDIT: Also, when it comes to sharpness, the lens has to resolve to the pixel density of the sensor. Although the GM1 is "only" 16MP, it has a pixel density of 7.11 MP/cm², where the 5DMKIII has a density of 2.58 MP/cm². The GM1 is actually a lot more demanding on the lens's resolving power even though it has 8less megapixels.
DXO is measuring resolution normalized to the frame, not absolute resolution or per-area resolution. i.e. it's measuring line pairs per image, not line pairs per mm. So, when comparing m43 with APS-C, it really does seem to come down to sensor area and pixel pitch. The high pixel pitch on m43 means that each pixel is tougher to resolve and so an APS-C pixel gets a higher per-pixel score than m43. So, even with equivalent MP between the 2, you'd expect APS-C to score higher. Since the sensor area is still going up faster than the MP count on APS-C their score ultimately gets a boost on both sides since the higher pixel count contributes as well.
the most illuminating comparison might be the Sigma lenses. On m43 they're somewhat pedestrian performers compared to native m43 lenses, but put the "exact same" lens on a Nex or A6000 and suddenly it's a stellar performer - the 'mpix' score goes up significantly even though the optical performance doesn't.
The lesson is that the good native m43 lenses really are exceptional, but the choice to go with a smaller sensor area does limit the ultimate absolute resolution that is achievable. In particular the new Oly Pro zooms seem to perform about as well as any zoom available on the Nex/Alpha APS-C systems despite the handicap of the smaller sensor.
Oops, should follow up with 2 additional tangential thoughts.
First, given the DXO scores it should be clear why adding more MP to the m43 sensors isn't a high priority - the sensor is already at such a fine pixel pitch that adding more MP to the same area is not likely to improve overall performance.
Second, this illustrates why the rumored/alleged 'sensor shift' approach in the E-M5MkII is interesting - it's a way to enhance the captured resolution without altering pixel pitch. this means that it doesn't make any further demands on the lens.
DXO is rating in RAW-files only..
Dxomark does publish their methodology.
top level index
I published my blogpost on that topic...
DXO gives a good baseline, but it's far from being a practical guide. For example they rate the 36mp sensor in the D800/10, A7R as the best. Yet they don't take into account it's practicality in real world use and/or metering abilities of the camera, lenses, etc. It's all technical based. I think in general bigger sensors are potentially better, but for example the lenses and smaller size of M43 has practical merit in real world use which DXO could never measure. I say that as an owner of both FF DSLRs and M43 and using both on a regular basis. It's better to base your camera purchases on the supporting system and not just the sensor.
This is the way that I use DxO.
I use it to compare lenses against each other within the same system, which is generally what I'm doing. I might want to know the technical difference between the Oly 12-40/2.8 and the Pany 12-35/2.8.
There are ways to get information that makes sense cross platform, but in all honesty it's not really worth the setup and at the end of the day - shoot with the system you like. Right now, I can't tell you a camera system made from 2004 and onward that falls down in general IQ. they are all really good. The big differences are in the things that you will care about more like AF speed, camera operation speed.
It's a long and complicated debate, with most things coming down to subjective items that mean more to you than me or vice versa.
I'd just say skip DxO for the most part, save yourself the headache.
Separate names with a comma.