Open Letter ...... ??????

[bernhardAS]

I am really puzzled by the "open letter to the Camera Manufacturers"

The kind of information and the topic of the Letter do seem to match the standards of contributions that are usual on LL.

1) The issue has been noted and fixed. What exactly has still to be done?
2) There is no explanation whatshowever how the third graph has been constructed. And no basis explanation or measurement that the alledged increase in ISO is actually happening.The basic problem is, that the sensor has some light loss that is not in the same amount "seen" by the light meter. The transmission loss of the lens is irrelevant as long as it is pichked up by the light meter in the same degree as the sensor. A software correction in known situations where the issue occurs, could be to "tune" the light meter output. A second possibility could be, to increase the gain on the sensor. Frankly the first alternative sound much easier to me.  Thwere is no explanation whatshoweevr how the conclusion is derived that the ISO gain is used for correction. Did someone disassemble the source code ?
3) The assumption that DoF is impacted by the issue rests also on very weak arguments and is as admitted in the article not based on any measurement.

I also may be a stickler on legal formalities, but the published graphs should in my opinion not really be marked as DXO confidential. If they are indeed confidential they should probably not be published.

I can  think of a number of other issues that are actually relevant and might warrant an open letter if one gets sufficiently agitated, but I can't see any reason for this article. 





 

[PetterStahre]

I couldn't agree more. The open letter got me confused. What's the actual problem to be solved?

Seems logical that glass/plastics/coatings/whatever in front of the CMOS/CCD interferes with light.
Seems logical that the camera manufacturer has to deal with that in software/hardware to produce readings according to standard ISO measurements. (And the open letter suggests they already are.)

How well they deal with it at any given ISO setting, compared to the competition, is measured on a daily basis on countless sites like this under topics like noise, dynamic range, color accuracy ...

What am I missing?

 

[pstoppani]

Indeed, an odd letter.

I find it interesting that the cameras for which they "show" automatic sensor gain aren't exactly the type of camera who's owners are likely to spend the money on expensive fast lenses.  The light loss for higher end cameras seems "in the noise" and no auto-gain data is given for these cameras.  So, my conclusion is: there is no problem worth worrying about here...

Also, how exactly did they measure the auto-gain in the sensor?  And where is the data for relevant cameras?

[bradleygibson]

I have no reason to distrust DxO's numbers--measuring auto-gain is fairly straightforward when the sensor's saturation point is known and the light intensity is controlled.

But I too am puzzled by the letter.  Is Mark saying that f/1.2 lenses absorb more light wide open than they do when stopped down?  (for example, f/1.2 = t/1.4, but f/2 = t/2.1?)  Or is he saying that wide open, lenses transmit relatively more off-axis light which the sensor is inefficient at capturing?  Either way, within reason, I'm glad that a stop is (approximately) a stop is a stop.  It would be a PITA to go from f/4 to f/2 in aperture, make the same compensation in the shutter speed and have a significantly different exposure.

For me?  Thanks but, no, thank you.  If, due to the laws of physics, we have to tweak the ISO a bit, I don't think that's such a bad solution.  A nice tech note buried somewhere in the back of the manual would be nice, but I'd be interested to hear how this would materially impact someone.  We've all been shooting for some time now, and it seems no one noticed...

Shutter speeds aren't completely accurate at the extremes, lens planes of focus shift as they are stopped down, gain is applied unequally to color channels, etc.  But these machines still get the job done, and done very well, on the whole.

Perhaps there's something I'm missing?

[bjanes]

But I too am puzzled by the letter.  Is Mark saying that f/1.2 lenses absorb more light wide open than they do when stopped down?  (for example, f/1.2 = t/1.4, but f/2 = t/2.1?)  Or is he saying that wide open, lenses transmit relatively more off-axis light which the sensor is inefficient at capturing? 

Bradley,

I think that the latter is the point of his letter. He confuses the subject by bringing in T-stops, which are determined by absorption of light by the lens elements and do not vary with the aperture. The former is corrected by TTL (through the lens metering), but the latter would complicate TTL.

[denopa]

The real point of the letter is that the wider a lens opens, the more light is lost. Since you pay (a lot) more for a fast lens, it's very important to understand that you get a lot less bang (light ?) for your buck on digital than on film, and even less on some cameras. You do get other things for the extra $$ (construction and focus speed mostly) but I'd venture that speed and DOF are the main attractions for most people - at least they are for me.
So my reaction is quite the opposite of yours, this is actually a very important "find", and well worth publishing on this website. if this could nudge manufacturers towards "smarter" sensors a la M9, it would be a very good result.

[inissila]

> you get a lot less bang (light ?) for your buck

Huh? 0.35 stops (e.g. D200 with f/1.4 lens) is "a lot"? 1/3 stops increase in gain results in a very difficult to notice difference in noise (when the discussion is e.g. between f/1.4 and f/2.8 lenses, 0.35 stops is not very interesting).

And whopping 0.1 stops ISO increase at f/2... this is just the daftest article I've read in a while.

[MarkL]

Of all the issues there are with current cameras - I don't think this even registers on the scale of significance.

[JBerardi]

Of all the issues there are with current cameras - I don't think this even registers on the scale of significance.

I'd agree. In a world where nearly every camera still uses a JPEG-based histogram, I'm not losing much sleep over .1 stops of ISO gain...

[uaiomex]

"The future belongs to Sony and Panasonic"
LuLa is on a rampage.

Eduardo

[BernardLanguillier]

The main point of the letter is probably the confirmation that LL believes in the relevance of DxO measurements.

Cheers,
Bernard

[michael]

It's unfortunate that some folk's reading skills (including the ability to interpret graphs) seems to have deteriorated somewhat.

Carry on.

Michael

Ps: As for "the future belonging to Sony and Panasonic", maybe a bit hyperbolic, but a grab of 30% marketshare for their mirrorless cameras in just a few years is really upsetting Canon and Nikon. Trust me!

[01af]

Either way, within reason, I'm glad that a stop is (approximately) a stop is a stop.

It would really be nice if things were so easy.

It would be a PITA to go from f/4 to f/2 in aperture, make the same compensation in the shutter speed and have a significantly different exposure.

As a matter of fact, this is exactly what you get—different exposures, that is—with most lenses when one of the apertures involved is the lens' maximum aperture. It's called the r-stop error.

A nice tech note buried somewhere in the back of the manual would be nice, but I'd be interested to hear how this would materially impact someone.

It would materially impact most. People are too lazy to read tech notes buried somewhere in the backs of manuals, and those who still do are too dumb to understand. Back in the '60s, '70s, and early '80s, Minolta Rokkor lenses had a mechanical r-stop error compensation. As a consequence, people complained and accused Minolta to lie about their lenses' real speeds, as according to the in-camera light meter, an f/1.4 lens looked more like f/1.7 or f/1.8 obviously. In fact, at that time Minolta was the only company that gave accurate exposures at all apertures but people didn't like it. Minolta tried to explain in their literature but it didn't help. It was too complicated to understand the technical underpinnings, so people preferred under-exposed pictures instead. When in the mid-'80s Minolta introduced the new auto-focus system with the new A bayonet mount (now Sony Alpha) and new lenses, they gave up r-stop error compensation, and everybody was happy from then on.

We've all been shooting for some time now, and it seems no one noticed ...

When there is some r-stop error compensation going on, either mechanically (as in the days of the Rokkor lenses) or electronically (as it would be possible today), people will notice, they won't understand, they will complain, and they will force the manufacturer to stop it or they gonna switch to Canon.

[uaiomex]

These figures are a surprise to me but as with any new fresh product, sales can grow exponentially, then they stabilze. Once the market is right, Canikon will show up with some killer mirrorless aps cameras.
However, I truly love LuLa being a "provocateur".
Eduardo

Michael

Ps: As for "the future belonging to Sony and Panasonic", maybe a bit hyperbolic, but a grab of 30% marketshare for their mirrorless cameras in just a few years is really upsetting Canon and Nikon. Trust me!
[/quote]

[captzoom]

It seems to me the corrections Fuji engineers, which they are calling "Sensor Optimization", are doing for the X100 is an indirect admission of the problem hypothized by your "Open Letter...".  They admit to the problem digital sensors have gathering light towards the edge of the senor.  Their section on how they developed the sensor and the lens for the X100 is worth exploring.

-CaptZoom

PS Your investigations, discoveries and insights are greatly appreciated.  Please continue the good work you guys do here at LuLa.