CCD Advantage

[Bill Pierce]

Here is an interesting piece which, among other things, points out the advantage and necessity with wide angle lenses of the CCD sensor in digital M Leicas as compared to the CMOS sensor in many DSLR's.


http://www.luminous-landscape.com/essays/an_open_letter_to_the_major_camera_manufacturers.shtml

 

[johnnygulliver]

a very interesting and thought provoking article Bill, many thanks for posting the link.

 

[antiquark]

http://www.luminous-landscape.com/essays/an_open_letter_to_the_major_camera_manufacturers.shtml

I'm not quite sure what the writer is saying, but it sounds like he doesn't understand how ISO works. Is he saying that a D700 should produce brighter pictures than a D40 (at the same ISO) because the sensor is better?

 

[semilog]

That is a fascinating and thoughtful article.

Moreover, it is extremely funny if you think about all the baying for "fast, wide primes" for various digital systems (especially micro 4/3). Apparently, super-fast lenses not only aren't that much brighter, but in some cases they won't even buy you shallower DOF!

Now, that said, the "CCD advantage" in the M8/9 has almost nothing to do with CCD structure (as implied but NOT demonstrated with data in the linked article). Rather, the advantage of the M8/9 comes from the use of an eccentric microlens array. In fact, the requirement for that specialized array is specifically because the CCD sensor prefers to see light entering normal (perpendicular) to the sensor plane.

Note that an eccentric microlens array is also being used by Fuji on the APS-C, CMOS sensor for the X-100 — which is fitted with a fast (f/2) wide angle lens.

Fuji knows what they're doing.

For BOTH CCD and CMOS sensors, the best approach will eventually be to use backside-illuminated sensors (back-illuminated CCDs have been common in scientific imaging for well over a decade, and back-illuminated CMOS sensors are increasingly prevalent in consumer devices with small sensors). A good article that shows how beneficial backside illumination can be is here (scroll down to Figure 3).

By putting the photosites closer to the sensor surface, the "tunnel effect" is minimised and quantum efficiency can be nearly perfect (approaching 100%).

Unfortunately, making large back-illuminated sensors is still very expensive. Too expensive, at present, for 4/3, APS-C, and larger sensors. And these sensors, because they are very very thin, are (mechanically) fragile — and the bigger they get, the more fragile they are. Not rugged enough to put in a Nikon D4, yet. But maybe by the time the D5 is out...

 

[semilog]

I'm not quite sure what the writer is saying, but it sounds like he doesn't understand how ISO works. Is he saying that a D700 should produce brighter pictures than a D40 (at the same ISO) because the sensor is better?

He clearly understands how ISO works, and that is not what he's saying.

 

[antiquark]

I re-read it... he's saying that the ISO is cranked up for wider apertures. That means that the camera would have to query the lens about the aperture, then adjust the amplifier gain accordingly.

A simple test of this theory would be to use a lens that doesn't communicate its aperture to the camera, and check if wide aperture images are darker than expected.

However, it would be a huge flaw if camera makers didn't normalize the ISO for wide apertures. Basically, every wide aperture photo would be underexposed. I can't see how people would perceive this as a "feature."

 

[semilog]

However, it would be a huge flaw if camera makers didn't normalize the ISO for wide apertures. Basically, every wide aperture photo would be underexposed. I can't see how people would perceive this as a "feature."

Of course that's right. People will complain, regardless. They want to have their f/1.2 cake and eat it, too. 🙄

But this also provides a rather vivid demonstration of the fact that RAW files and EXIF data don't always tell the "raw truth."

 

[eddie1960]

I re-read it... he's saying that the ISO is cranked up for wider apertures. That means that the camera would have to query the lens about the aperture, then adjust the amplifier gain accordingly.

A simple test of this theory would be to use a lens that doesn't communicate its aperture to the camera, and check if wide aperture images are darker than expected.

However, it would be a huge flaw if camera makers didn't normalize the ISO for wide apertures. Basically, every wide aperture photo would be underexposed. I can't see how people would perceive this as a "feature."

On Pentax cameras there are some metering exposure issues with k and m series lenses ( i have 4 i frequently use) and typically they are out about a third to a half stop

 

[Soothsayerman]

Here's an interesting article on CCD tech vs CMOS tech...

http://www.dalsa.com/corp/markets/ccd_vs_cmos.aspx

 

[semilog]

Here's an interesting article on CCD tech vs CMOS tech...

http://www.dalsa.com/corp/markets/ccd_vs_cmos.aspx

^---- solid summary.

 

[antiquark]

In the article he says: "One might be better off purchasing smaller aperture lenses and increasing the ISO."

However, if my calculations are correct (big if), you're still benefiting from fast lenses. The worst EV adjustment for f1.4 is 0.5, which means that the ISO will have to go from, say, 100 to 140. If you choose an f2.0 lens, you'll have to increase the ISO to 200, which will be noisier than ISO 140. Thus, the 1.4 lens isn't exactly a waste of money.

 

[semilog]

In the article he says: "One might be better off purchasing smaller aperture lenses and increasing the ISO."

However, if my calculations are correct (big if), you're still benefiting from fast lenses. The worst EV adjustment for f1.4 is 0.5, which means that the ISO will have to go from, say, 100 to 140. If you choose an f2.0 lens, you'll have to increase the ISO to 200, which will be noisier than ISO 140. Thus, the 1.4 lens isn't exactly a waste of money.

Assuming, of course, that the T stops of both lenses (lens T-stops, not sensor T stops) are actually proportional to their marked f/ numbers. That's not always a valid assumption, especially when you're comparing a 5 or 6 element f/2 lens to a 10 or 12 element f/1.2 lens.

Note that when people working in filmmaking buy a lens for $10, 20, 30k they expect that the aperture ring will show T-stops!

 

[semilog]

A major issue with CCD sensors on a camera that isn't sealed, i.e. fixed lens, is dust.

Don't remind me. All our (scientific) cameras are Peltier-cooled, so to prevent condensation they are under vacuum. We just lost a vacuum seal on our best EMCCD camera... :bang:

 

[JPSuisse]

This is really a very technical issue here. Although I'm an engineer and can even follow part of these very interesting articles (luminous landscapes and dalsa.com), I'm disappointed that we are stooping so low here. Or, that we have to go to such lengths to understand what big corporations are doing to us....

"The morality of art consists of the perfect use of an imperfect medium." The problem with all these in camera modifications is that the character of your lenses is probably being hidden. In such a way, you'll never master your imperfect medium.

For me, it's another reason not to replace my failing D100 with a new DSLR. I hate deception.

Kind regards,

JP

 

[Sonnar Brian]

In the end, you pick a digital camera that gives you the image that you want. With a film camera, you can just load it with the film that gives the look you want. Digital camera- stuck with the sensor. Comparing images from my M8 with the EP2, shooting both at Low ISO, 160 and 200 respectively, the biggest difference that I see is more noise in the shadow areas on the EP2 images. The M8 images are smoother, less noisy. CCD vs CMOS? Olympus boosting gain because of sensor vignetting? More signal processing on the M8- I know that is not true! They are fairly minimalist on the in-camera processing. I don't care- the M8 images look better.

The EP2 is fun, makes a good image, uses lots of different lenses, and I can use it for making movies. But the M8 images are "cleaner".

 

[JPSuisse]

PKR,

don't call me a hypocrite now...

Good thing I don't have to pay my bills with the equipment I've got. I can view it as pure art. The bills get paid by making semiconductor components for you guys buying this stuff! 🙂))

Now where's that NEOPAN 1600...

JP

 

[j j]

Can someone explain why any of this is a problem at all in any way? Even if true is it not just a camera correcting exposure?

 

[semilog]

Do you have some kind of vacuum pressure alarm.. ? What Torr is the sensor under?

No alarm -- it's a sealed assembly. We realised there was a problem when sensor performance plunged, it wouldn't hold temp at -80 (there's a thermocouple), and then we found condensation on the inside of the faceplate... :bang:

 

[Sonnar Brian]

No one knows who Jack Kilby or Fred Terman were. All they know is "Steve Jobs", a marketing guy..

I was using some 7400 series logic today.

 

[jrv]

If the author in the first article is upset to learn the camera applies a gain (ISO change) to normalize for different aperture settings, I wonder what they think of what the in-camera imagine processing is doing when a high-ISO setting is used? They might think "colorization" tame by comparison.

I would be very surprised if any camera CCD sensor has only a single readout channel. There are simply too many pixels, read too quickly, for ~ 14 bits of A/D. Uniformity is an issue for camera CCDs, but there are fewer places "drift" is introduced and calibration/correction should be easier than with CMOS.