[rawr]

Note he says "proved", not "guessed", and puts specific numbers to it. He is looking at the actual data, measuring standard deviations and other statistical stuff that goes way over my head (although back when I was a math major in college, I'm sure I'd have followed it better!).

Thanks for finding a relevant GBG quote on the issue. I've read it before but lost track of it in the discussion on DPR.

At the risk of bringing down a bolt of lightning on my head, GBG is still guessing, or more accurately observing, forming a hypothesis, testing it and then putting forward a theory. His work is very informed and based on a series of careful observations and tests he and others have made, but his observations merely indicate correlation, not proof.

That's the problem with reverse engineering a black-box. You see some results in the outputs of the black box that match your hypothesis and pass your tests, but you have no way of really knowing what caused the results within the black box because you can't open it up and look inside. If you can't look inside you have no real way of knowing whether the results you observe were intended, or co-incidental, or or a by-product of something else, or even produced by something else entirely within the black box that you didn't know was there.

You could only call the matter provenif your observations of the outputs of the black box matched the settings you saw within the machinery of the black box once you opened it and had a look inside. Bingo! You can now prove that the black box has been setup to output X in exactly the way you observe output X. Cause is clear and precisely matches effect. Case closed.

Statistically you could also probably consider the matter close enough 'proved' if you had enough observations from lots of K20D and K7 cameras to demonstrate an extremely high incidence of behaviour that supports your hypothesis, but that's hard to organise given the millions of cameras out there. So measuring results from a small sample won't get you anywhere near even statistical proof.

Phew!

Luckily pedantry is perfectly acceptable, even expected, on pentaxforums, otherwise I would have felt really awkward posting that.

[jgredline]

Thanks for finding a relevant GBG quote on the issue. I've read it before but lost track of it in the discussion on DPR.

At the risk of bringing down a bolt of lightning on my head, GBG is still guessing, or more accurately observing, forming a hypothesis, testing it and then putting forward a theory. His work is very informed and based on a series of careful observations and tests he and others have made, but his observations merely indicate correlation, not proof.

That's the problem with reverse engineering a black-box. You see some results in the outputs of the black box that match your hypothesis and pass your tests, but you have no way of really knowing what caused the results within the black box because you can't open it up and look inside. If you can't look inside you have no real way of knowing whether the results you observe were intended, or co-incidental, or or a by-product of something else, or even produced by something else entirely within the black box that you didn't know was there.

You could only call the matter provenif your observations of the outputs of the black box matched the settings you saw within the machinery of the black box once you opened it and had a look inside. Bingo! You can now prove that the black box has been setup to output X in exactly the way you observe output X. Cause is clear and precisely matches effect. Case closed.

Statistically you could also probably consider the matter close enough 'proved' if you had enough observations from lots of K20D and K7 cameras to demonstrate an extremely high incidence of behaviour that supports your hypothesis, but that's hard to organise given the millions of cameras out there. So measuring results from a small sample won't get you anywhere near even statistical proof.

Phew!

Luckily pedantry is perfectly acceptable, even expected, on pentaxforums, otherwise I would have felt really awkward posting that.

First, I would like to chime in on this and say that this is a most enjoyable conversation. I have really enjoyed reading your exchange with Mark.

I tend to agree with you here. Now as far as the highly regarded GBG fellow goes, I have no idea who this fellow is as I don't spend time at DPR other than to check out new things and their side by side deal. But even then he is also guessing in from what I read regarding the link provided.

Now here is my Question. It seems like all the talk has been around RAW images, but since I am a Jpeg shooter, there is no doubt in my mind the K20D produces better high ISO jpegs out of the box. Quite good actually. Why would this be?

Would it also not be easier to call Pentax and ask them about the Noise and if it is handled differently?

[nostatic]

Now here is my Question. It seems like all the talk has been around RAW images, but since I am a Jpeg shooter, there is no doubt in my mind the K20D produces better high ISO jpegs out of the box. Quite good actually. Why would this be?

Would it also not be easier to call Pentax and ask them about the Noise and if it is handled differently?

Have you tried *every* possible combination of settings on the K7? You cannot assume any parity in settings between K7 and K20d, and sometimes interactions of settings can give unpredictable results.

If by "out of the box" you mean just shooting what was set at the factory, then to me that has very little relevance. iirc, the K7 was set "vivid" from the factory and personally I think that setting on any camera I've used looks horrible. But I'm not into over-saturated colors and "hey look at my highlights" type of images.

[lastdodobird]

Taken with a K100D + DA40/2.8

ISO3200, f5.6

[Rupert]

I see all the technical talk, and marvel at the knowledge of others in such areas, but for me I just appreciate what I often get with my K20D at high ISO settings. Shots I might otherwise miss, or not even attempt. Perfect? By no means, but very usable and even print extremely well...even at large sizes. I just appreciate the shots, and that is good enough for me to be well pleased with my K20D, high or low ISO, it performs nicely for my needs.
K20D +DA*50-135 @135mm 1/30 Handheld F2.8 ISO 5000 Light by nearby porchlight.

https://www.pentaxforums.com/gallery/images/141/1_IMGP4754-800.jpg

BTW- How do I get the photo to appear a instead of just the link?
Regards!
Rupert

[jgredline]

Have you tried *every* possible combination of settings on the K7? You cannot assume any parity in settings between K7 and K20d, and sometimes interactions of settings can give unpredictable results.

No I have not. I have however set up both My K20D's (I have 2) and the K-7 to pretty much the same settings, as close as I could anyway. Both MY k20d's are pretty much the same, but the K-7 seems to be about 1/2 stop behind on ISO performance. Don't get me wrong, I am happy with the results of the K-7, but in order to get the same results as the K20D, I have to run it at lower ISO settings. This is not a problem for me, since I shoot outdoors in the sun 95% of the time. My questions are more out of curiosity as to why.

If by "out of the box" you mean just shooting what was set at the factory, then to me that has very little relevance. iirc, the K7 was set "vivid" from the factory and personally I think that setting on any camera I've used looks horrible. But I'm not into over-saturated colors and "hey look at my highlights" type of images.

I agree. I do like vivid mode for flowers and landscapes though, I shoot little of them.

[Gooshin]

https://www.pentaxforums.com/gallery/images/141/1_IMGP4754-800.jpg

BTW- How do I get the photo to appear a instead of just the link?
Regards!
Rupert

this forum allows HTML tags

to post your photo, you have to embed it in "code"

the start code is [IMG] and the end code it [/IMG]

put the link between these two tags, with no spaces, viola.

[Marc Sabatella]

At the risk of bringing down a bolt of lightning on my head, GBG is still guessing, or more accurately observing, forming a hypothesis, testing it and then putting forward a theory. His work is very informed and based on a series of careful observations and tests he and others have made, but his observations merely indicate correlation, not proof.

As I said, I don't pretend to have worked through the math to be able to say it doesn't constitute proof when he says it does. If you've worked through the math and have come to a different conclusion than he has, that would I'm sure be interesting discussion between you and him. But until then, I consider the matter as closed as it is ever likely to get.

[Marc Sabatella]

Now as far as the highly regarded GBG fellow goes, I have no idea who this fellow is as I don't spend time at DPR other than to check out new things and their side by side deal. But even then he is also guessing in from what I read regarding the link provided.

Again, not "guessing" at all. He perform statistical analysis on the RAW data itself. Noise is easily measured statistically, using something long the lines of standard deviations of the individual pixel values. And since we're starting with the same actual sensor data at any ISO level but simply amplifying it differently before the A/D conversion, it goes up in a very predictable way as you turn up ISO. If you see noise not going up by as much from ISO 800 to ISO 1600 as it should, then some sort of processing has gone on. It's really as simple as that. The sensor can't magically become less noisy at ISO 1600 - it's the same sensor capturing the same photon no matter what the ISO. What you call this processing is up to you - if you'r uncomfortable calling it noise reduction, fine, call it "George" if that makes you happy. But whatever you call what the K20D does, the K20D does it and the K-7 doesn't, and the evidence of this is right there in the data.

Now here is my Question. It seems like all the talk has been around RAW images, but since I am a Jpeg shooter, there is no doubt in my mind the K20D produces better high ISO jpegs out of the box. Quite good actually. Why would this be?

Well, I'd guess both the K-7 and K20D use more or less the same RAW->JPEG conversion. So if th k20D is doing some NR (or "Georging", if you prefer) on the RAW data, it start out with less noise before the conversion to JPEG, and ends up with less noise after the conversion to JPEG. Meaning you'd need to use a higher NR setting in your custom options to get the same results, unless Pentax tweaked the K-7 JPEG conversions algorithms to do more NR at each setting than it did on the K20D. But assuming you really have done back to back controlled comparisons using the same scene, same exposure, and same settings, and seen more noise in K20D JPEG's, then I guess that suggests you *do* have to turn up the JPEG NR option to a higher setting in order to make up for the lack of "Georging" before the conversion.

Given how much superior it is to do your NR in PP starting from RAW, comparing the default JPEG's seems kind of pointless, though. Kind of like comparing the highway mileage of two cars but only after putting tire chains on them. If you really care about gas mileage, why drive on the highway with tire chains? And if you really care about noise, why shoot JPEG and rely only on the in-camera NR? But if you must, then turning the K-7 up to a higher NR setting would probably put them on more even footing.

Would it also not be easier to call Pentax and ask them about the Noise and if it is handled differently?

What do you the chances would be of someone answering the phone who actually knows?

[lastdodobird]

Does increasing the EV help with high iso shots?

I did these test shots a few hours ago (DA40/2.8 + K100D @ ISO3200, f5.6) with +0.7EV and they seem to come out better...

[Ron Boggs]

Recently needed to do quicky images for daily press releases to newspapers and websites. Shot JPEG's at the lowest resolution available in the K10d (2 megapixels I believe) and at 1600 iso. Don't know anything about noise reduction so submitted to the papers without editing other than minor cropping and perhaps a bit of exposure correction. Ended up having several images published over a weeks time by as many as 7 different newspapers (for one of the images) so extremely low quality high iso cruddy ol' JPEG's apparently aren't too bad from Pentax digital cameras. Note that I had never previously gone beyond 400 iso nor ever shot at less than max megapixels...nor had I ever done any sports photography, so the "experiment" worked!

[jgredline]

Does ASA 1600 count?
ME SUPER, 50F/1.4, FUJI ASA 1600 SUPERIA

[Ron Boggs]

I'm stuck in the old film mode--400 is fast film. So for digi, 400 is fast to me. Had no clue that 1600 was even viewable, though I still consider it unprintable--the digital variety anyway, film is it's own beast. No reason not to shoot plain ol' family vacation photos that will never be printed at 1600 and low res. ----fit about a billion on a high capacity SD card!

I guess on second thought, even 1600 must be printable as it looks good up to 1/4 page in the newsprint version. Probably not for fine art or large prints, but for snapshot work (plus internet and newspapers)--low res JPEG's at high iso are now an option for me.

[arbib]

ISO 1600 / Sigma 70-300 DG APO at under 5'. and near the 300 end. Very slight crop.
Pentax K20d / No Flash / NR on Weakest in camera, no other NR done. Just the basic Post

[octavmandru]

I see all the technical talk, and marvel at the knowledge of others in such areas, but for me I just appreciate what I often get with my K20D at high ISO settings. Shots I might otherwise miss, or not even attempt. Perfect? By no means, but very usable and even print extremely well...even at large sizes. I just appreciate the shots, and that is good enough for me to be well pleased with my K20D, high or low ISO, it performs nicely for my needs.
K20D +DA*50-135 @135mm 1/30 Handheld F2.8 ISO 5000 Light by nearby porchlight.

https://www.pentaxforums.com/gallery/images/141/1_IMGP4754-800.jpg

BTW- How do I get the photo to appear a instead of just the link?
Regards!
Rupert

Just hit the yellow icon with a mountain, then paste the link inthe small window.

Does increasing the EV help with high iso shots?

I did these test shots a few hours ago (DA40/2.8 + K100D @ ISO3200, f5.6) with +0.7EV and they seem to come out better...

They do! I posted some results previously in this thread, Jpegs from Raw.
The chroma noise impact is drastically reduced, especially in dark areas when overexposing a bit in Raw and then pull it back when converting to Jpeg.

https://www.pentaxforums.com/forums/pentax-dslr-discussion/64295-pentax-high-...st-here-7.html