View from 30 000 feet

[Andrew Teakle]

Michael,

Thanks for your thoughtful and insightful article. It's a little sad to see the demise of so many "intitutions" of the imaging industry at such an exciting time in its evolutionary progress. I guess that's been going on in most industries as society and technology advances.

I applaud the apparent efforts of camera manufacturers to recently focus on "better" rather than just "more" pixels in their products. Having said this, I am slightly disappointed with the meagre upgrade of the 20D while appreciating the new features introduced. I guess I am exactly that consumer you referred to who expects revolutionary changes in camera specs every 18 months or so.

While you mention that the race for more megapixels is plateauing out, I read with interest a posting on DP Review last December that mentioned a 'revolutionally new chip". This CMOS sensor apparenly uses less power, gives a significant increase in dynamic range, and uses fewer transistors per pixel. My (probably incorrect) extrapoltion of the last point is that, as the area per pixel that is involved in gathering light is increasing (in proportion to the pixel size), pixel sizes will decrease as transistor sizes reduce while maintaining the same light-gathering characteristics. Surely this will lead to greater pixel densities with similar noise characteristics. With reduced power consumption, heat generation may also be reduced (again, maybe an unfounded deduction), reducing dark-field noise.

I don't have your wealth of experience at watching evolving technologies mature, and these new chips may be a decade or more away from being used in production cameras. I do, however, hope that chips with reduced power, higher dynamic range, smaller size, lower noise and (as in the article) faster processing are not too far away.

While I love the tools I currently use (5D & 20D, RZ67) I look forward to all advances in this exciting field. I just hope lens technology keeps pace.

Andrew

[skysnake]

  The inflight service would be great if he was on American Airlines. I couldn't resist the opportunity to toot my own airline's horn! LOL

Seriously though, Michael made a great quote in that article.

"All of the brand zealots that hang out on the forums need to take a pill"

So true. It literally drives me crazy that we get into this Ford/Chevy Canon/Nikon debate. It gets so emotional over at dpreview that fellow photographers literally HATE eachother. I just don't get it. We're all in this together. Having a little fun is one thing but when the brand zealots start hating eachother, it's gone too far. Just my $.02

Cheers!

[michael]

I've learned to never say never.

But, the laws of physics won't be easily challenged.

Once pixels get below about 5 microns it seems that it becomes almost impossible to get really decent signal to noise ratios. There simply aren't enough photons captured. The wells are too small.

Remember, we're actually starting to get close to just a few multiples of the wavelengths of visible light at these small sizes.

When I interviewed sensor development scientists at both Kodak and Dalsa last year they both agreed (off the record) that 5 microns was likely to be the cut off point for the forseeable future. At 5.5 microns the Nikon D2x shows this to be the case. Image quailty is very good, but starts to suffer the higher the ISO setting.

Since most lenses are running out of steam at this point, there isn't much point in pushing the envelope further.

There are a few digicams with sensors down to 3 microns, and the S/N is just awful, even at low ISO.


Michael

[Graeme Nattress]

I think sensor technology breakthroughs will happen that will re-write the equations of pixel size / dynamic range / noise, but they're probably a number of years away. An awful lot of research is going on in this area. And I don't know how far they'll be able to shrink pixel size and still get these advantages either.

However, one thing that is good about the ~35mm format size is the control over DOF it offers - you don't need an incredibly fast lens to get some nice out of focus backgrounds, and you don't need to stop it down too far to get everything sharp. It offers a kind of sweet spot.

Graeme

[Quentin]

Maybe its time to decamp to Birmingham (England that is   ) where I have just atended Focus on Imaging, a huge annual show that takes place at or around the same time as PMA.

I say this because both Hasselblad and Mamiya had a significant presence at the show.  Mamiya ZD's were positively sprinkled about the place (on three different stands).  Actually it was the importers, not Mamiya themselves, who had the formal presence, which is what you would expect, and the only ones with batteries in working ZDs.  I took the chance to stick my card in a ZD and shot a couple of raw files.  The ZD is the medium format Kodak SLR/n, in my opinion.  Big, a little Kludgy but fundamentally works just like a regular dslr.  Poor screen and viewfinder let it down a bit.  You'll either love it or hate it, probably both at the same time.  The files were highly detailed, but I (or the camera, with which I am not familiar) underexposed the shots, taken handheld at 125 ISO when base is 50 ISO.  Some noise therefore, but lots of detail and excellent colours out of both ACR 3.3 and SilkyPix.  Said to be available in  "a couple of weeks"....The samples they had were going to mags for testing.

The Hassy H2D was demo'ed by a top fashion photographer whoise name I have forgotten but who produced some stunning work.  Leaf and Phase One also had there wares on show.

I managed to fondle a Nikon 105mm F2.8 Macro VR sample - seemed to me VR worked at 1:1 to a degree.  The new Carl Zeiss F-mount lenses were also available as samples on the Robert White stand - excellent build quality, and they seem to meter OK - no autofocus of course.

These were a few of my personal highlights.

Quentin

[BJL]

... I read with interest a posting on DP Review last December that mentioned a 'revolutionally new chip". This CMOS sensor apparenly uses less power, gives a significant increase in dynamic range, and uses fewer transistors per pixel.[a href=\"index.php?act=findpost&pid=59337\"][{POST_SNAPBACK}][/a]

Some of that sounds like what Panasonic has already done with its nMOS sensors, which reduce the number of transistors to two per pixel, from about four per pixel in other CMOS sensors. That was launched last year in tiny camera phone sensors, and is now used in the "Live MOS" 4/3 format sensor of the Olympus E-330 and Panasonic Lumix DMC-L1.

I am more optimistic about the dynamic range of smaller pixels than Michael. At minimum ISO, some compact digicam sensors seem to give fairly good DR, more than enough for scenes of up to somewhat above normal subject brightness range, and they now do it with pixel spacing of under 3 microns. For example, going from the fairly good DR sat minimum ISO of some digicams with the 3.4 micron pixels of 5MP 2/3" CCD's, I can see 3.4 microns being quite usable as far as DR goes for low shutter speed/good light/high resolution work, giving over 20MP even in 4/3 format.

I expect that the upper limits on pixel counts will come more from some combination of  the resolution that lenses and the print resolutions for which there is sufficient demand.

Let me predict somewhat mainstream DSLRs going past 10MP (D200, Pentax this fall ...), some vaguely affordable ones going past 20MP (1DsMkIII?, D3X?? ...) and only an handful of extremely expensive high end models going past 30MP (MF only?).

[John Camp]

This PMA may eventually regarded as a kind of negative revolution, in the sense there was nothing new that any rational person might argue is a "gotta have." Maybe we are on a plateau; maybe a 1DsII and a D2x will be seen as perfectly good shooters five years from now, in a way that the Nikon 1D  (which came out when; 2000?) no longer is. If that's the case, then maybe some of the outside camera companies, like Pentax and Leica, will get a little breathing space, and be able to get their footing back. I saw somewhere a very good review of a Zeiss lens on, I believe, a Canon 1DsII with an adapter. The reviewer claimed a substantial benefit over L lenses of the same length. So maybe the competition will begin to move to glass, or to something other than megapixels.

I have a question that I haven't been able to get answered on any Nikon forum, and that is, could you use Nikon DX lenses with a 1.33 crop, like the Leica DMR has, as opposed to the Nikon's current 1.5? Does Nikon have room to build that much larger a sensor and still retain usability in the DX line?

One thing that I'll be watching Luminous Landscape for is a review of the new printers from non-Epson sources. I don't have a problem with Epsons, but it'd be nice to have enough competition that we get some downward pressure on Epson ink pricing. You'd think their ink is bird flu vaccine,  judging from the prices. So -- some printer reviews would be great.

JC

[Ray]

I think perhaps a little basic maths is appropriate here. The 20D has 6.4 micron pixels, the D2X 5.5 micron pixels and the D30, if it were a 12mp camera, would have had 5.2 micron pixels. All are within that 5 micron limit that Michael mentions.

It seems to be Canon's policy that they don't release models that have worse noise in any respect than a previous model. My guess is, a 5.2 micron 12mp D30, at this stage of development, would have had worse noise at ISO 1600 and 3200 than the 20D, and that that wasn't acceptable. A way around this would have been for them to have removed the ISO 1600 option, as they did with the D60, but that also would have been unacceptable now people are used to the luxury of relatively noise-free performance at high ISOs. The 5D has raised the bar in this respect.

We should also not forget that the 8mp of the 20D translates to 20mp on a full frame sensor, without reduction of pixel size, and the 12mp that everyone was hoping the 30D would sport, translates to 30mp on a 24x36mm sensor.

[giles]

I have a question that I haven't been able to get answered on any Nikon forum, and that is, could you use Nikon DX lenses with a 1.33 crop, like the Leica DMR has, as opposed to the Nikon's current 1.5? Does Nikon have room to build that much larger a sensor and still retain usability in the DX line?
[{POST_SNAPBACK}][/a]

[a href=\"http://bythom.com]Thom Hogan[/url] has tested some DX lenses on full frame cameras.  I am dubious that any manufacturer will have consistently created 1.3x coverage on "DX" type lenses, but you never know, I suppose.  Quoting from Thom's Nikon 12-24mm review:

That "wow!" is even more interesting in that the 12-24mm can be used on a 35mm body from about 18-24mm. At 18mm there is a clear light falloff at the corners, but at 24mm this is no worse than the 24mm f/2.8 Nikkor at the same apertures.


Giles

[bob mccarthy]

Thom Hogan has tested some DX lenses on full frame cameras.  I am dubious that any manufacturer will have consistently created 1.3x coverage on "DX" type lenses, but you never know, I suppose.  Quoting from Thom's Nikon 12-24mm review:

That "wow!" is even more interesting in that the 12-24mm can be used on a 35mm body from about 18-24mm. At 18mm there is a clear light falloff at the corners, but at 24mm this is no worse than the 24mm f/2.8 Nikkor at the same apertures.


Giles
[a href=\"index.php?act=findpost&pid=59443\"][{POST_SNAPBACK}][/a]

The same is true of the 17-55/f2.8. I can use it on the F5 at the longer focal lengths - roughly 28mm up.

Bob

[BJL]

.... the 12-24mm can be used on a 35mm body from about 18-24mm. At 18mm there is a clear light falloff at the corners, but at 24mm this is no worse than the 24mm f/2.8 Nikkor at the same apertures.
[a href=\"index.php?act=findpost&pid=59443\"][{POST_SNAPBACK}][/a]

No surprise: zooming in with a zoom lens typically works by effectively magnifying the image given by its shortest focal length, with that image then cropped by the lens barrel, lens mount, and the film or sensor. Zooming in from 12mm to 18mm is a 1.5x magnification, just enough to magnify the image from DX frame size to 35mm frame size. And when you go all the way to 24mm, that zoom magnification is enough to get rid of the corner light fall-off seen at 18mm.
[Added: the example of the 17-55/2.8 works exactly the same way: zooming from 17mm to 28mm gives 1.5x magnification again, matching the format factor again.]

(Aside: This is a cousin of the optical fact that at focal lengths of about 50mm and up, almost any lens designed to fit a 35mm format lens mount will produce an image circle that covers then 35mm frame, even if the design is optimized for a smaller format like DX.)

On the other hand, the fact that at 12mm the image is just big enough to cover the DX format frame, and not a just slightly larger 1.3x one, is yet another little piece of evidence that Nikon is in no way preparing to reverse direction and make a liar of itself by increasing its DSLR sensor sizes.

To borrow Michael's phrase, those of you who are stuck on 20th century film formats as timeless universal ideals, and imagine seeing evidence for a return to them almost everywhere you look, need to Get Over It!

[jani]

Some of that sounds like what Panasonic has already done with its nMOS sensors, which reduce the number of transistors to two per pixel, from about four per pixel in other CMOS sensors.

Yes, this is pretty interesting (and possibly so obscure that it might be off-topic).

If I recall correctly from when I took classes in circuit design:

The well-functioning compromise was to design CMOS circuits of nMOS and pMOS transistors, where the nMOS transistors took care of pull-down between output and low-voltage while pMOS transistors handled pull-up between output and high-voltage. In CMOS, nMOS transistors are off when pMOS transistors are on, and vice versa. This is fairly efficient, because nMOS is slower when pulling up than pulling down. CMOS also had advantages in relation to noise, etc.

That nMOS is now making a comeback is fascinating; it's over forty years since CMOS appeared on the scene, and over twenty since nMOS "died". Does this mean that they've found a way to make sure that the transistors mostly perform pull-down of voltage, hence playing on the strengths of nMOS?

If I am to speculate, it may have to do with how photons charge the photosites, and maybe they've also utilized the more modern nMOS processes' lower voltage threshold.

[Brian Gilkes]

One thing that I'll be watching Luminous Landscape for is a review of the new printers from non-Epson sources. I don't have a problem with Epsons, but it'd be nice to have enough competition that we get some downward pressure on Epson ink pricing. You'd think their ink is bird flu vaccine,  judging from the prices. So -- some printer reviews would be great.

JC

I think Michael has one on the new Canon 12 colour jobs coming up.
It's not only ink cost that needs addressing, but speed, software (including RIPs), and ability to autofeed sheets into 24" and larger printers amongst other things.
We need better yellow inks ,  K5 carbons included in the inkset and of course no matte black /photoblack change problem. The latter hassle may have seen it's last expression in the current Epson x800 printer series.
I think we may see more happening here than in increased pixels in cameras.
Roll on Photokina.
Cheers,
Brian
www.pharoseditions.com.au