[justinr]

Can't say that I've read the whole of this forum but it seems that everybody thinks that a FF camera would replace the K7. Why should this be so? The 5D a 1D live happily alongside one another as well as their smaller stable mates.

I run a Mamiya ZD for portrait and studio work and a K10 where a dSLR is required and to be quite honest I am contentiously surprised at just by how much I can crop the smaller cameras image and still retain the quality. However, it will never match the ZD for the depth and sense of clarity that the Mamiya produces. Would a FF Pentax go beyond the K series and start to threaten my love affair with dMF? It will be interesting to see.

[Class A]

I'd say low noise is much, much more useful to most photog than being able to print a 45x30".

And what's the relationship between one and the other?

More pixels don't mean more noise -> More pixels offsets noise!.

[Miserere]

It's my impression that Pentax Japan is very much the "real" company, with the various international affiliates responsible for marketing, distribution, and contracting for warranty service in their respective countries. There does not seem to be much flow of technical information in either direction.

(This has been a particular frustration in my effort to put solid, detailed facts in my flash guide, but I think it's reasonable to generalize.)

You're quite right, Matt. It would be nice if what the international sales outlets had to say was taken seriously by Pentax HQ, but I don't think they care. This is pretty standard procedure for Japanese companies. Then again, do American companies care what users in Japan have to say or do they listen to their American clients first and foremost?


...

[24X36NOW]

And what's the relationship between one and the other?

More pixels don't mean more noise -> More pixels offsets noise!.

The relationship is higher pixel density = more noise.

The article you referenced simply points out that to reduce the noise of the higher pixel density camera to approach that of a lower pixel density camera, you can downsample the high pixel density to a lower pixel density. In other words, the only way to reduce the noise is to lower the resolution to that of the lower resolution camera, so the relationship is only "offset" by (more or less) equalizing pixel density.

In other words, there's no such thing as free lunch.

[Class A]

The relationship is higher pixel density = more noise.

"More noise" where?
If you look at a single pixel, yes, it will be noisier.

If you compare the images at 100%, yes, the higher pixel density image will look noiser.

However, that's not a fair comparison, is it?

Or do you think that my K100D yields sharper images than a K-7? According to a 100% view comparison, it would yield sharper images because when I look at the K100D 6MP image at 100% it looks a lot sharper than the 14.7MP image.

A 6MP image will always have more contrast between adjacent pixels than a 14.7MP image (everything else being equal and excluding extreme test patterns). That's doesn't make the 6MP sensor the more acute one.

I hope the resolution example makes it clear that comparing images with different resolution (size) at 100% doesn't make sense.

A fair image comparison is performed on images with the same size. That either means downsampling the larger image or upsampling the smaller image (or variations thereof, such as downsampling with different degrees; printing them to the same size would be a practical method).

In other words, the only way to reduce the noise is to lower the resolution to that of the lower resolution camera,

Note that downsampling a higher resolution image will not only reduce noise but also retain more detail of the scene, compared to the lower resolution image.

In other words, there's no such thing as free lunch.

I guess we weren't talking about "lunch" then.

Your noise argument only applies when you compare images at their largest possible output size (for a given dpi boundary). A higher pixel density image can be printed larger and if you do it to the extent that the pixels become as big as the one in the lower resolution image then, yes, you'll see more noise. But this is comparing apples with oranges.

[Winder]

The relationship is higher pixel density = more noise.

The article you referenced simply points out that to reduce the noise of the higher pixel density camera to approach that of a lower pixel density camera, you can downsample the high pixel density to a lower pixel density. In other words, the only way to reduce the noise is to lower the resolution to that of the lower resolution camera, so the relationship is only "offset" by (more or less) equalizing pixel density.

In other words, there's no such thing as free lunch.

Actually you can increase pixel density and improve the s/n ratio at the same time. Canon has done it repeatedly.

The Sony A900 has 2x the photosites as my Canon 5D, but the s/n ratio is almost the same. The Sony A900 is actually slightly better than my 12MP 5D. Since the photosites (pixels) are half the size the "noise" appears as much finer grain to the viewer, so even if the s/n ratio stays the same and you increase the MP you will get a cleaner looking image.

The quality of RAW converters is improving as well, which is just as important.

[thibs]

Sony 10Mpix -> 12Mpix is another example: more pixels, less noise.

[er1kksen]

And of course Canon 40D->50D is a wonderful example of pixels increasing, noise decreasing, and the general public drastically mispercieving, with DPReview at the helm.

Of course, the 40D was a camera with so-so high ISO performance and a whole bunch of hype. The hype was clearly the important part, and when the 50D came out with objectively better high ISO performance, hype won the day nonetheless.

[jpzk]

Just give me:

1. Faster and more reliable AF "lock-on" for BIF shots
2. Improved high-ISO performance (read: noise handling),
3. Higher synch speed for flash,
and already I would be happy to consider a replacement for the K7.
I do like my K7 (and my K20D), don't get me wrong, but these two features certainly would make me do a lot of thinking.

JP

[Nimrad]

I would be happy with:
- Better sensor in general (better colors, sharpness etc)
- AWEsome ISO-performance
- compact body
- GPS with auto-tagging!!! If not in-body, at least a batterygrip!

[24X36NOW]

"More noise" where?
If you look at a single pixel, yes, it will be noisier.

If you compare the images at 100%, yes, the higher pixel density image will look noiser.

However, that's not a fair comparison, is it?

Or do you think that my K100D yields sharper images than a K-7? According to a 100% view comparison, it would yield sharper images because when I look at the K100D 6MP image at 100% it looks a lot sharper than the 14.7MP image.

A 6MP image will always have more contrast between adjacent pixels than a 14.7MP image (everything else being equal and excluding extreme test patterns). That's doesn't make the 6MP sensor the more acute one.

I hope the resolution example makes it clear that comparing images with different resolution (size) at 100% doesn't make sense.

A fair image comparison is performed on images with the same size. That either means downsampling the larger image or upsampling the smaller image (or variations thereof, such as downsampling with different degrees; printing them to the same size would be a practical method).


Note that downsampling a higher resolution image will not only reduce noise but also retain more detail of the scene, compared to the lower resolution image.


I guess we weren't talking about "lunch" then.

Your noise argument only applies when you compare images at their largest possible output size (for a given dpi boundary). A higher pixel density image can be printed larger and if you do it to the extent that the pixels become as big as the one in the lower resolution image then, yes, you'll see more noise. But this is comparing apples with oranges.

If I compare D300S with D3S, with slight resize of D300S to equalize pixel count (D300S actually has a slightly higher pixel count than D3S @ 12.3MP vs. 12.1MP), the D300S image is still MUCH noisier, as it has a higher pixel density (due to smaller format with similar pixel count). Maybe what you mean is higher pixel COUNT compensates for additional per pixel noise? Certainly higher pixel density does not, at least not in all cases. A D3 vs. D3X comparison doesn't seem to show much difference after a downres of the D3X file, but that may account for higher pixel COUNT as opposed to higher pixel DENSITY.

(All comparisons based on ISO 6400 using "still life" images from Image Resource saved and resized to the same (as close as possible) pixel counts.)

Also, if as a "practical matter" (i.e., as you put it, comparing prints of the same size) additional pixels would compensate for the additional noise, why don't we see a D3X with ISO 102,400??

[24X36NOW]

Actually you can increase pixel density and improve the s/n ratio at the same time. Canon has done it repeatedly.

The Sony A900 has 2x the photosites as my Canon 5D, but the s/n ratio is almost the same. The Sony A900 is actually slightly better than my 12MP 5D. Since the photosites (pixels) are half the size the "noise" appears as much finer grain to the viewer, so even if the s/n ratio stays the same and you increase the MP you will get a cleaner looking image.

The quality of RAW converters is improving as well, which is just as important.

Sony 10Mpix -> 12Mpix is another example: more pixels, less noise.

And of course Canon 40D->50D is a wonderful example of pixels increasing, noise decreasing, and the general public drastically mispercieving, with DPReview at the helm.

Of course, the 40D was a camera with so-so high ISO performance and a whole bunch of hype. The hype was clearly the important part, and when the 50D came out with objectively better high ISO performance, hype won the day nonetheless.

Comparisons of different generations of sensors would seem quite pointless; of course the signal to noise (among other things) is improved with each new generation of sensor, but that doesn't really shed any light on the pixel density vs. noise, which can only be studied in a meaningful way with sensors with similar technology.

[er1kksen]

If I compare D300S with D3S, with slight resize of D300S to equalize pixel count (D300S actually has a slightly higher pixel count than D3S @ 12.3MP vs. 12.1MP), the D300S image is still MUCH noisier, as it has a higher pixel density (due to smaller format with similar pixel count). Maybe what you mean is higher pixel COUNT compensates for additional per pixel noise? Certainly higher pixel density does not, at least not in all cases. A D3 vs. D3X comparison doesn't seem to show much difference after a downres of the D3X file, but that may account for higher pixel COUNT as opposed to higher pixel DENSITY.

(All comparisons based on ISO 6400 using "still life" images from Image Resource saved and resized to the same (as close as possible) pixel counts.)

Also, if as a "practical matter" (i.e., as you put it, comparing prints of the same size) additional pixels would compensate for the additional noise, why don't we see a D3X with ISO 102,400??

Well, you see, there are TWO variables you've introduced by comparing the D3 and D300s. There's the pixel density, and there's the sensor size. Larger medium=less enlargement for equal output size. Noise characteristics are therefore less pronounced. Your comparison does not isolate pixel density as a variable.

Differences in image noise between equally sized sensors of the same tech generation have not correlated with pixel density. Some with more are cleaner, some with less are noisier. With microlenses preventing loss of light to nonactive sensel regions, four small pixels gather light just as efficiently as one larger pixel (and naturally sample noise more accurately as well, though equal-sized output will look the same or better).

The determining factors now are technology developments and sensor area. Pixel density is not a major factor now that the problem of non-active sensel areas has been solved.

The D3x doesn't have ISO 102,400 (neither does the D3) because it's a high-res studio, landscape, and advertising camera. Superhigh ISOs like the D3s features are not in demand in its niche. If you really want to shoot at that ISO with a D3x you can push the RAW files and probably get results about as good as you'd get pushing D3 files (D3s is better but it's a generation past the D3 and D3x).

Comparisons of different generations of sensors would seem quite pointless; of course the signal to noise (among other things) is improved with each new generation of sensor, but that doesn't really shed any light on the pixel density vs. noise, which can only be studied in a meaningful way with sensors with similar technology.

I agree, the example I stated was meant to highlight the public misperception of the issue, rather than to compare the 40D and 50D in any other way than to note that the 50D had lower noise, objectively (regardless of whether the reason was pixel density or sensor generation or anything else), and the mobs ignored the observable fact because their conspiracy theory that camera companies were increasing pixel densities just to screw them out of their money was just more fun(?) to believe in.

I actually see room for a new model:

New sensor - 8 FPS, ISO12800, EXT:25600
New AF system - 30% faster, improved low light capability, tracking
New flash system - faster, improved wireless capabilities
Improved liveview - hybrid focusing capability(contrast AF)
Improved video - HD Video, limited AF capability
Improved buffer - 35 RAW frame capacity @ 8 FPS
Dual card slots - Overflow and transfer capabilities, WIFI compliance

Announcement: Feb. 2011
Availability: May 2011

Behold the "Pentax K-8"

I'm open to wagers

OOOHH! an 8FPS Pentax Camera, bring it on! That would make Canon and Nikon sit up and take notice!

[rawr]

I still don't fully get the enthusiasm for FF I pick up here.

Maybe I need educating, but my take on FF is that it indisputably has virtues for low-light high ISO, and therefore whoever needs that needs FF. People who like or need their wides to be wides also need FF.

But FF doesn't also automatically provide the same scale of benefit for other important parameters of IQ as it does for low-light high ISO. Colour depth and dynamic range are hardly any different on FF as on crop, viz:


('Portrait' means Color Depth, 'Landscape' means dynamic range, 'Sports' means low-light high ISO)

So what am I getting if I move to FF, if not better IQ? Not much, it seems.

I will usually get a bigger, brighter viewfinder, which would be great, of course. But I will also get narrower DOF in landscapes and everywhere else, and lose the handy focal length multiplier of crop, especially useful for sports and birding. And FF is always way more expensive than crop.

FF is still not adding up for me.