[calsan]

I have 3 aps cameras.

Ordered by pixel size largest to smallest: K10d, K7, K3
Ordered by noise most to least: K10d, K7, K3.

Just saying...

[reh321]

I have 3 aps cameras.

Ordered by pixel size largest to smallest: K10d, K7, K3
Ordered by noise most to least: K10d, K7, K3.

Just saying...

OK, so you've just "proven" that each generation of cameras does a better job of handling noise.
(I thought that was an unspoken "given")

[bdery]

Kids, I thought I said enough :P

Because it's Friday, here's a gift:

Rainbow-Cake Recipe Inspires Comment Apocalypse

A very fine summary explanation, thanks, but you don't really expect this to change anything, do you?

I'm that naive.

P.S. Perhaps you already answered this, but...
From what I see you've been quite impressed with the 70-200. Do you intent to buy it?

It's a VERY impressive lens. Stellar performance and manufacturing from what I've seen. Plus it looks badass.

But... it's too big for my needs. The 60-250 was fine on the K20D, and on the K-3 it's magical. I don't need a shorter range and I don't need a faster lens. I can always use the DFA 100 macro f2.8 WR if needed. So no, I won't buy it. If anything, I'd be more likely to buy the 24-70 (or the 20-24).

"Equivalence in simple terms means trying to get the exact same image with two different formats.“ is a complete misrepresentation. The notion of "equivalence" is about a correct understanding of the impact of format size on image quality (noise and detail). While such a correct understanding allows one to figure out when images will be equivalent even when produced on different formats, it is inappropriate to equate a notion with one of its applications or, worse, claim that this application is the goal of having the notion.

I think you're playing with words, but it might be the language barrier on my side.

Noise is not driven by pixel size.

Yes it is. All else being equal, it is. Dark current is a MUCH more important contributor to noise than shot noise, which occurs mostly in photomultiplicators and avalanche photodiodes. It is visible when counting photons, which we are far from doing with our analog sensors.

The evidence exists on DxOMark

Mixing "evidence" and "DxO" in the same sentence is funny. With 1-element sample groups, infinite standard deviation, and biased tests, they are useless to this discussion.

Here is a horse that won't stop dying, no matter how much one beats it.

Optical designers, engineers and graduates in optics all beg to differ.

Yes, there are some highly technical second-order influences of pixel size on pixel noise, but the first-order effects and light shot-noise dominate and we are talking about image noise not pixel noise, hence one can forget about pixel size.

As I wrote, shot noise is a real issue when counting photons. We are NOT counting photons with our cameras.

I have 3 aps cameras.

Ordered by pixel size largest to smallest: K10d, K7, K3
Ordered by noise most to least: K10d, K7, K3.

All things are not being equal when you compare technology from 8 years apart.

[bertwert]

Because it's Friday, here's a gift:

Rainbow-Cake Recipe Inspires Comment Apocalypse

Oh god I can't stop laughing

So how long do you freeze the numbers for?

[Class A]

I think you're playing with words, but it might be the language barrier on my side.

Personally, if I thought that a language barrier may be preventing me from understanding what another person says, I would not insult them, just in case. Perhaps you should try that sometime.

Surely, it cannot escape you that people are agreeing with you based on the misleading aspects of your "summary" and that's why I think this is not a matter of "playing with words" but about being honest and precise what a proper understanding of format sizes is about.

I find your statement offensive and very much regret that you are not able to understand why your "summary" is a 100% inadequate. I guess that's why you came up with it in the first place, but I thought a bit of elaboration may make you see your mistake. It apparently didn't. My mistake.

Dark current is a MUCH more important contributor to noise than shot noise, which occurs mostly in photomultiplicators and avalanche photodiodes. It is visible when counting photons, which we are far from doing with our analog sensors.

When I use "shot noise", I refer to "photon shot noise", i.e., the quantum nature of light. Photons follow a Poisson distribution and hence cause image noise. This noise and first order noise issues on sensors dominate over the secondary effect that is influenced by pixel size.

Do you know that readout noise is actually lower for smaller pixels and that it does not add up linearly but with the square root of the number of pixels?

I only responded to your post because apparently it wasn't clear what I meant with "shot noise" before. I won't respond to your statements on pixel noise again. Anyone interested in the topic can check the article "Contrary to conventional wisdom, higher resolution actually compensates for noise, falconeye's article on "camera equivalence", or start a discussion with falconeye himself. I don't have any desire to engage in this fruitless discussion anymore.

Mixing "evidence" and "DxO" in the same sentence is funny. With 1-element sample groups, infinite standard deviation, and biased tests, they are useless to this discussion.

What is your proof for "biased tests"? What do you mean by "infinite standard deviation" in this context? What is your alternative to measurements that are respected by the industry? Your opinion?

As credentials seem to be important to you, note that the person behind Sensorgen.info -- a site that uses DxOMark data to compute sensor parameters -- is a Professor at the University of Wolverhampton. I guess you are saying he is an idiot to trust DxOMark?

Optical designers, engineers and graduates in optics all beg to differ.

"All" of them?
Or is it just you?

Can you point me to a place where they "beg to differ"?

BTW, the topic of sensor noise is not an "optics" topic. But perhaps try your luck and throw your optics credentials at Richard Butler from DPreview explaining to him why his article on "The effect of pixel size on noise" is so very wrong.

He presents empirical evidence to support his statements (in addition to the math), so perhaps, it would be just better to learn something and move on instead of claiming that everyone disagreeing with you on pixel size and its significance regarding image noise is wrong.

[bdery]

Personally, if I thought that a language barrier may be preventing me from understanding what another person says, I would not insult them, just in case. Perhaps you should try that sometime.

I don't believe I insulted anyone so far.

Photons follow a Poisson distribution

Only when the distribution is small enough. Small as in, as I said, "so small you can count individual photons". We are far from this when taking pictures.

"playing with words"

I wrote that because you refuse to understand that we are not talking about the same thing. I am talking about the noise on an individual pixel, not disagreeing with you. But you are NOT talking about what I AM talking.

apparently it wasn't clear what I meant with "shot noise"

You were clear enough for me, but you are making a mistake by assuming it's the prevalent source of noise when we have more than a few individual photons.

Anyone interested in the topic can check the article "Contrary to conventional wisdom, higher resolution actually compensates for noise, falconeye's article on "camera equivalence", or start a discussion with falconeye himself. I don't have any desire to engage in this fruitless discussion anymore.

It seems to me you are the one who won't let go here.

What do you mean by "infinite standard deviation" in this context?

Sample size of 1 = infinite standard deviation

What is your alternative to measurements that are respected by the industry? Your opinion?

Proper sample sizes, at least an average of 15 samples from varying production runs, by different operators. Ideally more than 15, depending on the level of accuracy you desire.

Assuming the observations obtained from 1 sample are perfectly representative of the whole product is simply wrong.

I guess you are saying he is an idiot

I never called anyone an idiot.

Richard Butler from DPreview explaining to him why his article on "The effect of pixel size on noise" is so very wrong.

He is discussing average noise over an area. I often wrote in agreement to this when people complained that the K-3's sensor had worse pixel-per-pixel noise performances than the K-5. However in this particular thread I am discussing noise at the pixel size.

In fact that's not what I was discussing at all at first. I was saying the 70-200 was too big for me.

[Class A]

I don't believe I insulted anyone so far.

You did not curse at anyone but I found the implication of your "playing with words" statement insulting and I have told you that I found it offensive.

The only way in this situation for you not having insulted me is for the insult to be a figment of my imagination. Here's why I don't think we are dealing with the latter case:

"Playing with words" at the very least implies that my statements had no substance. If someone has real arguments to put forward, why would they be "playing with words" instead?

Furthermore, why do people "play with words"? Typically, they have manipulation in mind, don't they? They don't have real arguments but still want to achieve an effect. Often, people "playing with words", are trying to weasel their way out of a losing argument without admitting to the fact that they have nothing left to counter with.

None of this applied in my case. I provided a solid explanation why your summary of what "equivalence" is about was misleading. I'd even go as far as saying it was a clearly written explanation, so I take issue with someone referring to my factual explanation as "playing with words".

This discussion is getting a bit out of hand. I did feel uneasy somewhat venting my frustration in a response to you in the first place. Other people said much more offensive things ("This nonsense again", "What drivel", ...), but I found it hard to maintain an entirely friendly tone when responding to your post, partially because I expect more from someone who studied optics.

Only when the distribution is small enough. Small as in, as I said, "so small you can count individual photons". We are far from this when taking pictures.

That's not true.

If you have so many photons that we are "far from" counting "individual photons" then shot noise absolutely dominates. Modern sensors have a read out noise of just a few electrons, so in the presence of a lot of photons, read out noise becomes insignificant.

Anyhow, may I remind you that the noise argument started with your claim that "Noise is driven (all other things being equal) by the pixel size." and that this is unequivocally wrong (when talking about "image noise" as I did) and I already gave you a lot of pointers. No point in shifting the discussion away from the original points and no point going around in circles either.

I wrote that because you refuse to understand that we are not talking about the same thing. I am talking about the noise on an individual pixel, not disagreeing with you. But you are NOT talking about what I AM talking.

OK, fair enough. It would help tremendously if you explicitly stated that you are talking about "pixel noise" when you use the term "noise". In particular, when responding to a post of mine, in which I explicitly talked about "image noise".

If you just use "noise" I think it is very reasonable to assume that you are talking about something that photographers should have an interest in, i.e., "image noise".

Out of curiosity, why are you concerned with the noise of an individual pixel?
If you understand that as the number of pixels increase the pixel noise increases as well (everything else being equal) but the overall noise does not, then why do you care about pixel noise?

Sample size of 1 = infinite standard deviation

You could also argue that the variance is zero and that (biased) standard deviation is the square root of the variance.

Anyhow, the point is that increasing the number of samples when measuring sensors would make it harder for an outlier to influence the results, however, I don't understand why this concerns you. Do you believe there is high variance in sensor production? Surely, if the theory on noise and DxO measurements coincide, you don't suspect that this is due to an unfortunate coincidence of sensor outliers, or do you?

The main point is that the question of how accurately DxO measurements represent the average performance of 15 or more sensors is entirely irrelevant. Your dissatisfaction with DxO's precision does not invalidate the argument I made (which was about pixel size not being a significant influence on image noise). If you reject DxO as evidence based on a "one sample" approach then you are saying that the agreement between theory and DxO measurements is purely due to sample variation. That's an untenable position, AFAIC, not forgetting that one can find other evidence from other sources.

BTW, DxO measurements confirm your statement that pixel noise increases with decreasing pixel size. Does that give you more confidence in their results?

Assuming the observations obtained from 1 sample are perfectly representative of the whole product is simply wrong.

I never stated that DxO achieves a "perfect" representation. This is a strawman.

I never called anyone an idiot.

Not explicitly, but if you state that "Mixing "evidence" and "DxO" in the same sentence is funny." you are implying that it is foolish to rely on DxO measurements. By implication, it would be idiotic to create a website whose results are based on DxO measurements, if it really were true that DxO results cannot be trusted.

He is discussing average noise over an area. I often wrote in agreement to this when people complained that the K-3's sensor had worse pixel-per-pixel noise performances than the K-5. However in this particular thread I am discussing noise at the pixel size.

Again, can you please let me know why you are talking about the noise of an individual pixel?

The only way in which pixel noise would be relevant to a photographer is if the photographer made the size of prints (or monitor viewing sizes) dependent on the number of source pixels. In practice, however, photographers use certain output sizes independently of the number of source pixels. An 8x10 print does not become an 16x20 print just because someone went from 6MP to 24MP.

It is true that when one tries to exploit the increased resolution of a higher MP image by printing larger that one will then see more noise, but enlargement always causes apparent noise to increase. I have always made it clear that I'm talking about image noise, so when you are suddenly discussing pixel noise (and not make the shift explicitly clear), it is obvious that your arguments do not make sense to me.

In fact that's not what I was discussing at all at first. I was saying the 70-200 was too big for me.

Now, if only you just said that.

I can guarantee you that neither Mistral75 nor I would have responded to a simple "too big for me" statement. Instead you wrote --

Seeing the 60-250 next to the 70-200 convinced me that for my lifestyle I don't want FF.

-- implying that format size had a significant impact on lens size/weight.

Now, you then agreed that Mistral75's and my responses were valid and everything would have been fine if someone had not picked up on the format size comparison again. I specifically and on multiple occasions suggested to not enter this discussion but it did not help. In the end, I'm characterised as someone who "shoves the equivalence discussion into people's faces". Oh, well.

---

Last edited by Class A; 11-24-2015 at 03:57 AM.

[rawr]

Only the bravest should post their personal report about a visit to a camera show on PentaxForums. ...

[bertwert]

Only the bravest should post their personal report about a visit to a camera show on PentaxForums. ...

Just don't go...

[crewl1]

Only the bravest should post their personal report about a visit to a camera show on PentaxForums. ...

Wait is that what this thread was about?

[K David]

I thought the only source of noise in a Pentax camera was the AF.

[monochrome]

Someone should come up with a Sony camera knockoff product for them. Just a shell with no guts inside and sell it for only half the price of the real deal.
Maybe this is how I'l finally strike it rich.

RE: Fashion cameras - Sony is the middle class hipster's Leica then

[Wired]

wait so my camera is powered by dark matter and it fires photon torpedo's through the lens? beam me up scotty!

[Riggomatic]

I'm hungry for cake now.