virtualrain
Offline
• • •
Upload & Sell: Off
|
 p.3 #11 · DxOmark: FE 100-400mm GM, compact and optically excellent | |
RCicala wrote:
We tested it, thought the results were accurate, but Sony insisted they couldn't be, they knew the lens was better than we showed.
Given that the lens had new technology, we were completely willing to consider our results might not be accurate and tried several suggestions from Sony, and several of our own ideas. Then we waited and repeated tests on later batches in case just the early ones had problems. Our results were always the same and as we investigated the lens more we found multiple reasons to be more comfortable with our results. They are valid and accurate.
There is still the possibility that the lens performs better close up than at infinity, but that always exists. But the Sony computer generated results that are way better than our real world tests are done at infinity. So a lot of time was spent confirming that the lens isn't as good in real life as Sony thought it was going to be.
Doing all this took many months. By the time we were done we figured not many people cared and we had lots of other things to do, so I never wrote up an official report. And there's the part that Sony fanboys are becoming the worst and writing a negative Sony article means hours and hours of responding to emails and comments.
...Show more →
---------------------------------------------
Brandon Dube wrote:
I get to have multiple opinions, both as something of an expert on metrology, and also as a lens designer. I think the 70-200 GM has a flawed design for a telephoto zoom lens. Aspheres are not generally all that helpful in zoom lenses unless you put them very close to an entrance or exit pupil, or at the very front. The 70-200 GM has three(!) two of which have huge aspheric departures, all very close together, and some part of moving groups if memory serves.
As the beam walks around on the asphere during zoom, it will change from being the ideal shape to potentially a very bad shape. Obviously this is not notionally the case here, since the lens is very good as-designed. However, I suspect those aspheres are hellishly sensitive to misalignment.
The lens also has to orchestrate two focus motions, and there is no feedback circuit. There's a LUT in the lens, like any other, for the position of the USM motor to reach a given focusing distance. The PDAF/CDAF system spits out an estimate for the differential between where the lens thinks it is, and where the AF wants it to go. The microcontroller looks up this position in the LUT for large defocus, and goes there. As you near focus, the control may change from master-slave to a loop where the motor drives and they are looking to maximize some focus metric.
In this lens, the near focus control simply changes over to the second motor. If the USM motor is in the wrong place, the image quality will suffer and this is not correctable within their control loop. I think this is principally the issue with this lens, and why the IQ is not up to expectation.
The optical design and "system engnineering" (AF, IS, opto-mechanical links) contribute as much to variability as manufacturing tolerances and QC. It's really easy to design a lens that's great on paper, but very difficult to build because it is sensitive to misalignment. The real challenge is a high performance, insensitive design....Show more →
Thank you both for sharing this... It all makes perfect sense. As others have said, we value your testing and analysis greatly.
|
.
