[ivanp]

...The guy in video...

Or perhaps he said something completely different than I think - actually I never learned English (I am Czech)

[PentaxPoke]

I think it is misunderstanding. The guy in video does not claim that EDR leads to higher noise, but that with activated EDR "the minimum ISO is increased to 200 and it may give more noise in shadow details".

As I understand it, you can either
compare apples to apples = pictures with ISO 200+EDR OFF and ISO 200+EDR ON (and the noise will be the same),
or compare apples to oranges = pictures with ISO 100+EDR OFF and ISO 200+EDR ON (and the noise will be different, but not because of EDR, but because using different ISO speeds).

I disagree. It depends on how Pentax implements the feature. No one but Pentax seems to know. One theory is that when you set the camera to EDR On and ISO 200, the camera will take the picture at 100 ISO and "push" the exposure to near iso 200 but stop short of blowing a highlight. Even though the effective ISO is approximately the same, the shadow noise is higher because it is noisier to push an exposure than to actually take the exposure at the higher ISO. According to this theory, this is why minimum ISO is 200 with EDR on since setting the ISO 100 is the base and lowest ISO setting for the sensor. If they were to allow 100 ISO setting with EDR on, then the camera would have to take the picture at ISO 50, which is not possible.

BTW: your English is just fine.

[WMBP]

I think it is misunderstanding. The guy in video does not claim that EDR leads to higher noise, but that with activated EDR "the minimum ISO is increased to 200 and it may give more noise in shadow details".

As I understand it, you can either
compare apples to apples = pictures with ISO 200+EDR OFF and ISO 200+EDR ON (and the noise will be the same),
or compare apples to oranges = pictures with ISO 100+EDR OFF and ISO 200+EDR ON (and the noise will be different, but not because of EDR, but because using different ISO speeds).

ivanp's observation is a good one. I'm not sure he's right, but it's something I hadn't thought of. Perhaps EDR increases noise simply because it boosts the ISO by one stop.

But gosh, there HAS to be more to it than that, no? I mean, there's got to be something going on besides the elimination of ISO 100. And if EDR does (a) cause the shot to be somewhat underexposed so that (b) the camera can then pull up the shadows, well, then I'd expect an increase in noise. AT least that's always the case when I post-process shots on the computer. Pulling detail out of underexposed shadows always results in more noise than I'd have gotten if I'd exposed those details right in the first place.

By the way, I'm having trouble understanding why EDR eliminates ISO 100. I think what has been said here is that, if you set your ISO at (say) 200, the K20D actually exposes the shot at ISO 100 and then does some in-camera boosting of the signal (if that's the right term) by one stop. Is that about right? But does that in itself reduce clipping?

Will

[PentaxPoke]

Will,

You and I may have been replying at the same time, but in my last post, I put forth one theory about why the limitation is 200 iso.

It is plausible that it exposes at 1 stop lower and then pushes back up but not quite all the way back up to avoid clipping.

[ivanp]

Manual says
By using the Expand Dynamic Range function, you can expand the light
level expressed by the CMOS sensor pixels, making it more difficult for
bright areas to occur in the image.
Which again leads me to convinction that if I set camera to 200 ISO, then it really shoots using 200 ISO, but too bright areas are "saved" by locally decreasing ISO up to 100.

[flyer]

Where can I read that information ? In the manual ?

I dont remember where I read the information. It was shortly after the K20D was introduced, and it was a transcript of a dialog between a journalist and a Pentax guy. It was clearly stated that the EDR was not "on" by default because it increased noise, notably in shadows. To me, it doesn't mean it doesn't have any possible use, it's just another tool that might be handy in certain circumstances.

[WMBP]

Which again leads me to convinction that if I set camera to 200 ISO, then it really shoots using 200 ISO, but too bright areas are "saved" by locally decreasing ISO up to 100.

OK, if that were the case - if there really is some sort of discrimination going on, causing parts of the photo to be exposed at different effect ISO's - well, that would be something a bit different from what I can do in post-processing.

But PentaxPoke's theory (that is, the theory that PentaxPoke has described) sounds more likely, because it's simpler, and it more clearly explains why there's no ISO when EDR is enabled.

Is there a reason Pentax Inc can't simply release a short tech note on this and spill the beans about what's really happening? Or do camera companies just not DO that sort of thing?

If it were really miraculous, I could understand that they might want to keep their secret to themselves. But it ain't miraculous.

Will

[PentaxPoke]

Manual says
By using the Expand Dynamic Range function, you can expand the light
level expressed by the CMOS sensor pixels, making it more difficult for
bright areas to occur in the image.
Which again leads me to convinction that if I set camera to 200 ISO, then it really shoots using 200 ISO, but too bright areas are "saved" by locally decreasing ISO up to 100.

Ivan, I can see both possibilities. In your theory, what you are really doing is adjusting contrast down so that the histogram is compressed because you are locally adjusting iso. I really don't see any histogram compression in the examples posted, including the latest here:
Pentax K20D review Cameralabs features: lenses

In the theory I am stating, the camera internally exposes to the left, and shifts the histogram back to the right. It will shift back to the right up to a 200% recovery of saturated highlights. If the highlights are beyond 200%, then they will still be blown.

Will, I think the reason they don't state what is going on is because they would essentially be revealing that they are not increasing Dynamic Range. Increasing dynamic range means the sensor would be capable of recording a wider range of light levels. I really don't think that is what is going on. IMO, it is a histogram-shifter-clipper-reducer, and that is consistent with every example I have seen.

[WMBP]

Will, I think the reason they don't state what is going on is because they would essentially be revealing that they are not increasing Dynamic Range. Increasing dynamic range means the sensor would be capable of recording a wider range of light levels. I really don't think that is what is going on. IMO, it is a histogram-shifter-clipper-reducer, and that is consistent with every example I have seen.

I agree, Poke, although it doesn't increase my respect for Pentax Inc. Another case of some dweeb in Marketing muddying the water with hyperbole, amazingly failing to realize that the company's market has a lot of technically minded people in it who are either aren't going to be fooled or at least going to be curious.

Will

[ivanp]

Ivan, I can see both possibilities. In your theory, what you are really doing is adjusting contrast down so that the histogram is compressed because you are locally adjusting iso. I really don't see any histogram compression in the examples posted, including the latest here:
Pentax K20D review Cameralabs features: lenses

Thanks for this great link, because the histograms confirm exactly what I am saying For easier comparison, I strongly recommend you to save both histograms to files, and compare them by quickly switching between both of them - you will clearly see that approximately 3/4 of both histograms are the same, while the right part with EDR is compressed.

[WMBP]

Thanks for this great link, because the histograms confirm exactly what I am saying. For easier comparison, I strongly recommend you to save both histograms to files, and compare them by quickly switching between both of them - you will clearly see that approximately 3/4 of both histograms are the same, while the right part with EDR is compressed.

I agree that those two histograms are interesting, and if I can find half an hour, I think I'll try doing something like this myself. The left three-fourths of the two histograms don't look quite identical to me, but pretty close; and I do see what ivanp is calling "compression" in the right-most quarter of the histogram.

So Ivan, your theory is that, with EDR on, the camera is shooting at the selected ISO (say, 200) for most of the dynamic range, but that it is somehow dropping the sensitivity to 100 for the highlights part of the histogram? Is that right?

Will

[PentaxPoke]

OK maybe I have been to hasty. At the suggestion of genesdn on the Flickr k20d group I ran a test that is similar to what he saw with the D300. Flickr: Discussing k20d Expanded Dynamic Range opinions in Pentax K20D / Samsung GX-20

Sure enough, the true value of EDR shows itself. It works on the RGB channels as well! So even if the brightness histogram we see in the viewfinder is not blown, one of the RGB channels could effectively be and so color information can be lost. EDR can help protect that.

All exposure values are exactly the same in these two shots. Both are RAW conversions. Note that the wagon is supposed to be RED!

EDR OFF



EDR ON



From now on, when I am shooting outside, EDR stays ON! Just another example of how much one can learn on these forums. Where else could we have found this information or suggestions for tests?

[WMBP]

Poke,

I'll take you out for barbecue next time you're in Dallas. You've shown me the first example of a pair of shots where I can see a truly significant difference between EDR off and EDR on. I haven't seen it this clearly even in my own tests.

And the link to the Flickr group was very interesting, as well - well, the link to Flickr, which includes the link to Ken Rockwell's site. I hadn't seen that explanation of how it works on the D300 before.

We still don't know that your earlier theory is wrong, do we? Or do you consider your test a conclusive refutation of your earlier theory?

Whatever the truth, I admit that your example is pretty compelling. And I will add - a bit retrospectively - that, while I didn't see any huge benefit to shooting with EDR on over the last couple of weekends, I didn't see any disadvantage at all. So perhaps you're right, keep it on while shooting outdoors in the bright sunlight.

THANKS!

Will

[PentaxPoke]

Thanks Will. I am just up I-35 in Oklahoma, so I may take you up on that.

I think what this proves is that I had no idea how EDR worked. but I am finally glad to see its true value.

[WMBP]

Thanks Will. I am just up I-35 in Oklahoma, so I may take you up on that. I think what this proves is that I had no idea how EDR worked. But I am finally glad to see its true value.

Poke,

Well, if you do come down, we'll go to Sonny Bryan's. Barbecue there is worth six hours in the car.

As for EDR, I still have no idea how it works. My question all along has been, does it do anything that I can't easily do better on the computer? From your example, I'd say the answer is a definite maybe.

Will