[Darley]

It mostly comes down to reduced sensor noise in the newer cameras I think.
Pixel count helps when the image is heavily cropped.
Also provided images are reasonably exposed, there is not much benefit from having a 16 bit sample depth in the camera while we still use 8 bit video hardware on the monitor.
And I wonder if the eye could resolve any better that 8 bit anyway?

The ist ds and K-01 both have 12 bits per sample so maximum brightness is Hex &H0FFF
I took a raw photo in the dark room with each camera grossly undereexposed so that the levels peaked at about &H003F
That is about 7 stops underexposed and means the resolution is limited to only 1 part in 127.
Also it is well down among the sensor noise. The raw images were completely black when opened in a viewer with 0EV of exp comp
Then used ImageMagick >convert -auto-level to automatically map the images up to normal brightness as 16 bit tiff, full brightness is &HFFFF
Then GIMP converted 16 bit tiff to 8 bit jpg for the video driver and monitors

The upper image is the istds and the lower is the K-01. The ist ds artifacts are mostly noise I think

Interesting - the K-01 shot appears brighter, sharper and more detailed but the *ist shot appears to be getting more detail in the shadows.

[Class A]

But to go back to the original question, fewer pixels does not mean less noise in the image.

Exactly.

A good article on that subject is the DxOMark article about the (absence of any) influence of pixel size on low-light performance.

Ah Dan but I'm not talking about noise, I'm talking about light sensitivity.

Light sensitivity pretty much equates to noise performance. In terms of quantum efficiency, most sensors are very close to each other; the main difference is how much noise they add on their own.

The sensor in the K100D (shared with the D40, as Dan mentioned) was a Sony sensor and was pretty good for its time. But the better column-parallel conversion architecture Sony used in the sensor for the K-5 is way superior. I have both cameras and there is just no contest. As has been mentioned before, the K-x was a really good low-light performer as well.

So would a modern 6MP sensor be better than a 12MP sensor and if so why don't they make them?!

No, it wouldn't be better.
That's the reason why they are not produced anymore.

BTW, if you downscale a 12MP image to 6MP then you get the same per-pixel noise performance as if you had shot with 6MP in the first place (provided the sensor technology is the same).

I was only going by Ken Rockwell's comparison of the 5D with APS sensors and the images from the 5D were clearly sharper and more detailed. Surely RAW files are not 'smoothed' in any way?

1. Avoid Ken Rockwell.
2. Avoid Ken Rockwell at all times.
3. Make sure any APS-C vs FF comparison is done fairly. If the same f-ratio (say f/2.8) is used for both, the FF shot will always look better. It uses more total amount of light, so the noise levels will be about a stop better. Larger sensors do not have an intrinsic low-light advantage. That is a myth and if you see a comparison that favours a larger sensor then either the larger sensor technology is better, or -- more likely -- the comparison has been done using the same shooting parameters (as opposed to the "equivalent" ones).
4. Some RAW data is smoothed. Unfortunately, Pentax belongs to the offenders. Even if you turn off all optional in-camera NR, Pentax still massages high-ISO RAW files to achieve better noise performance.

BTW, you can pretty much trust DxOMark results.
They do not look at spatial resolution when looking at high-ISO noise but then nobody does.
In other words, you won't find better measurements anywhere else and whatever RAW data massaging goes on is not that terrible that it invalidates the figures they report.

[Stone G.]

Some RAW data is smoothed. Unfortunately, Pentax belongs to the offenders. Even if you turn off all optional in-camera NR, Pentax still massages high-ISO RAW files to achieve better noise performance.

Isn't that true for all brands of DSLRs? In particular, with CMOS sensors with their individual drive- and read circuits for each and every pixel/photodiode, some smoothing has got to be a necessity before data is written into the raw files?

Apart from pixel size, sensor materials, electronics and "hidden" firmware smoothening there is one further aspect of sensor/pixel comparison, namely micro lensing that increases the amount of light that will actually hit the photodiodes. As far as I know, micro lensing technology has also undergone significant improvements over the years.

All in all, I firmly believe that while it may be relatively easy to compare images taken with different camera bodies and sensors, (be it 100% viewing or compressed to full screen viewing and be it with identical or equivalent lenses), it is quite hazardous to ascribe differences to one single element of one's entire system.

[Stone G.]

No, it wouldn't be better.
That's the reason why they are not produced anymore.

This is only partially true. Large-pixel sensors are still made, but not for consumer DSLR's but for special purpose cameras where long integration times and high dynamic range are of concern:

SBIG ST-402ME Class 1 Single Sensor Monochrome CCD Camera - OPT Telescopes

SBIG STF-8300 Color CCD Camera - OPT Telescopes

It is quite interesting to study specs stated here regarding Quantum efficiency, Full well capacity, pixel sizes, dark current......

....some information that we shall never see in our consumer DSLR manuals.

(And then, do note the prices!)

[ScooterMaxi Jim]

Isn't that true for all brands of DSLRs? In particular, with CMOS sensors with their individual drive- and read circuits for each and every pixel/photodiode, some smoothing has got to be a necessity before data is written into the raw files?

Apart from pixel size, sensor materials, electronics and "hidden" firmware smoothening there is one further aspect of sensor/pixel comparison, namely micro lensing that increases the amount of light that will actually hit the photodiodes. As far as I know, micro lensing technology has also undergone significant improvements over the years.

All in all, I firmly believe that while it may be relatively easy to compare images taken with different camera bodies and sensors, (be it 100% viewing or compressed to full screen viewing and be it with identical or equivalent lenses); it is quite hazardous to ascribe differences to one single element of one's entire system.

Good analysis. The micro lens challenge seems to be a very gradually improving situation, apparently as dependent on the quality of camera as date of production. I say that because the enlightening Dubovoy essay and DXOmark calculations (which unfortunately measured only Nikon and Canon bodies) regarding light acceptance T loss in fast lenses: An Open Letter To The Major Camera Manufacturers. Possibly the photo well has as much to do with it as the micro lens, I don't know the answer to that. Take note that the D70 - which had the same Sony sensor we are discussing on the Pentax units - was a poor performer, as was the D40x (10 mp) but not the D200 (also 10 mp but probably a higher spec version). Things got better by the time of the 12 mp units (D90/D300).

[Rondec]

Tech has changed so much since the K100 was new, that there is absolutely no benefit to getting a K100 if you have a kx or newer camera. A K100 could be shot at iso 1600, but had quite a bit of noise at that level -- certainly photos were much less usable than a K5 shot in same setting.

[ScooterMaxi Jim]

Tech has changed so much since the K100 was new, that there is absolutely no benefit to getting a K100 if you have a kx or newer camera. A K100 could be shot at iso 1600, but had quite a bit of noise at that level -- certainly photos were much less usable than a K5 shot in same setting.

Not that this thread has anywhere left to go, but the OP wanted to know - based on his 10 mp based camera - if there is any benefit to getting a cheap, older 6 mp for low light use. The answer should have been a simple "yes" and that would be the end of the discussion. If you get a KX or newer body for a price similar to any of the 6 mp bodies - of course, that would be the best possible deal. In theory - but that's an answer to a different question than what was posed.

[audiobomber]

Not that this thread has anywhere left to go, but the OP wanted to know - based on his 10 mp based camera - if there is any benefit to getting a cheap, older 6 mp for low light use. The answer should have been a simple "yes" and that would be the end of the discussion.

How about a slightly qualified "no"? There is no benefit to a 6mp Pentax camera for someone who has a 10mp Pentax camera, unless he is interested in the larger OVF and TTL flash on the *istD. And the advantages of having a second body of course.

The low light performance (SNR) of the 6mp bodies is exactly equal to the 10mp bodies.

[Class A]

How about a slightly qualified "no"?

Yes, to that.

[GeneV]

While much later bodies like the K5 especially and the K30 have far superior high ISO they are not cheap. The K10 and K200 share the same sensor and I was disappointed at the high ISO performance of Hegel K10 compared to my *istD. But other things like SR and higher resolution outweigh the high ISO disadvantage.

Same experience here with the 10mp K10d vs. 6mp K100d. There are lots of advantages to the K10d, but low light was not one of them.

[GeneV]

Tech has changed so much since the K100 was new, that there is absolutely no benefit to getting a K100 if you have a kx or newer camera. A K100 could be shot at iso 1600, but had quite a bit of noise at that level -- certainly photos were much less usable than a K5 shot in same setting.

I totally agree about the Kx. This was the first DSLR I owned where I felt halfway decent about even ISO 3200.

However, I took a K100 as a backup to my Kx on a trip three years ago because both used AA batts, and I was till not unhappy with the K100 shots.

[audiobomber]

DXOMark says the 6mp and 10mp sensor SNR's are the same. I've never heard anyone with deeper knowledge of the science of imaging (e.g. GordonBGood, Falconeyes, etc) criticize DXO test procedures, so I have to say that you guys who are seeing a difference in noise are mistaken.

[ScooterMaxi Jim]

When DXOmark does real world testing in the field, I'll take that as a big step forward. However, the majority of users of the two sensors who have commented here and at the other forum indicate that the rough look of the noise generated in the 10 mp sensors at high ISO is far more unappealing than the 6 mp noise (viewing at the same image size - not at the pixel level). I'm not aware of anything DXOmark evaluates accounting for subjective factors. Thing is - a lot of us don't accept numbers as the only measure when it comes to evaluating image quality.

I don't own a Limited lens - and they rarely rate high in numerical ratings - but their users feel pretty strongly (and the images are far more impressive than the rankings). Sometimes its just best to trust your eyes.

[ScooterMaxi Jim]

DXOMark says the 6mp and 10mp sensor SNR's are the same. I've never heard anyone with deeper knowledge of the science of imaging (e.g. GordonBGood, Falconeyes, etc) criticize DXO test procedures, so I have to say that you guys who are seeing a difference in noise are mistaken.

Which 6 mp Pentax dSLR was used in the comparison? It would be rather deceptive to make a reference to a comparison that simply does not actually exist in the context of the thread. I'm not aware that anyone here has brought Nikon into the discussion...

[audiobomber]

However, the majority of users of the two sensors who have commented here and at the other forum indicate that the rough look of the noise generated in the 10 mp sensors at high ISO is far more unappealing than the 6 mp noise

Not all have, and I've heard from lots of owners of both over the years who agree that there's no difference.

I don't own a Limited lens - and they rarely rate high in numerical ratings - but their users feel pretty strongly (and the images are far more impressive than the rankings). Sometimes its just best to trust your eyes.

You need to know exactly what they are rating. DXOMark measures MTF numbers, vignetting and CA, which are objective and inarguable. Then they assign a score, which attempts to describe the performance of the lens in a single number. The summary score is arguable, as are other factors which are not measurable (e.g. bokeh, build quality, etc).

Lots of people look at DXOMark lens ratings without realizing that the score measures the camera and lens in combination. The score from a lens measured on a full-frame camera will often be higher than a similar lens measured on an APS-C camera, which makes Pentax lenses look like they are lower performance. This is the source of most complaints of DXO lens scores.