I just don't get the whole "Expose to the Right" philosophy

[Barrie Davis]

But nobody can use higher ISOs as means for lowering noise, can they?

It depends. Clearly, the best way is to lower the shutter speed -- if you can. But let's say you can't due to motion blur or camera shake. Then, assuming, of course, that ETTR is possible for the scene (that is, you don't blow too much of the image by exposing more to the right than what the camera meters), you would up the ISO to ETTR rather than lower the shutter speed.

The reason this works is because higher ISOs have less apparent read noise than lower ISOs. The reason higher ISO images are more noisy is because they are made from less light, not because the sensor is more noisy at higher ISOs (since the exact opposite is true).

Hmm... R-i-g-h-t.... :-|

I'm gonna have to see if that can be made to work with my newest camera.. but it is not a dSLR. Has this read-noise improvement made its way to compact cameras, say, like my Panny LX3 I'm just learning to get the best from??
--
Regards,
Baz

I am 'Looking for Henry Lee ' (could be Lea, or even Leigh) and despite going 'Hey round the corner', and looking 'behind the bush', I have not yet found him. If he survives, Henry is in his mid-60s, British, and quite intellectual.

What is it all about? Well, something relating to a conversation we had in the pub 35 years ago has come to spectacular fruition, and I'd like him to know how right he was.

If you know somebody who could be this man, please put him in touch with me. Thank you.

 

[John Sheehy]

But think it through....

• To do EttR needs more exposure.

Absolute and relative (increasing exposure at same ISO), or just relative (raising ISO).

• More exposure means lower ISO is being used.

Yes; maybe not the setting, but the actual exposure index.

• Lower ISO always means less noise.... [doh!]

Not true. A low-key shot at ISO 125 will have more read noise than at 160 on most recent APS Canons and the 5D2, even when using the camera's metering. The 125 will have about 2/3 stop more read noise, and only 1/6 stop less shot noise. If we're talking absolute exposure being fixed, the lowest ISOs will add the most read noise to the image. It is a myth that high ISOs "add" more noise to an image. High ISOs are generally a cleaner operation on the camera, but with a smaller window of sensor charges digitized.

• Why not just use lower ISO to start with, and skip all the faffing about?

Because it may result in an open, soft lens, blur, or under-exposure with extra noise.

--
John

 

[GaborSch]

I would increase the intensity for example by Fill light, so that details of the xmas tree become visible and show the effect of increasing the ISO.

However, there is a problem if you can demonstrate, that ISO 3200 is better than ISO 100 +1 EV in PP, because the 5D does not have real ISO 3200.

--
Gabor

http://www.panopeeper.com/panorama/pano.htm

 

[John Sheehy]

Hmm... R-i-g-h-t.... :-|

I'm gonna have to see if that can be made to work with my newest camera.. but it is not a dSLR. Has this read-noise improvement made its way to compact cameras, say, like my Panny LX3 I'm just learning to get the best from??

No. The LX3 has read noise almost perfectly proportional to ISO.

The behavior referred to is only true of Canon and Nikon DSLRs, maybe some others, but I haven't seen any evidence. Some of the Nikons become almost proportional at lower ISOs than the other Nikon DSLRs. Older Nikons like the D2X have read noise very proportional to ISO (and quite high).

--
John

 

[John Sheehy]

I made the exact point that low ISO does not by itself lower absolute noise; but that it is the increased absolute exposure (that low ISO encourages) which does this. However AFAICT no camera has quite the neat and exact theoretical relationship of noise to signal across different ISOs, that you suggest. Especially if a flat linear curve is NOT used in conversion processing, as for example with camera JPG.

Actually, outside of Canon DSLRs, and some Nikons, digital cameras generally have read noise proportional to ISO. The gains for high ISOs on most P&S cameras are a joke; they just digitize noise. They'd be better off just using one gain, and moving the whitepoint around in the RAW data for ISO (just like many medium format digitals do).

--
John

 

[John Sheehy]

It is true that you haven't altered the film ISO rating as determined by sensitometric test at the point of manufacture, but you are EXPOSING it as if you had... which is the same thing when it comes to the shutter speeds and apertures employed at the point of hitting the button....

....and that is ALL THAT MATTERS!

Every time you increase the exposure over indicated by the meter, you are USING a lower ISO!!

It would simplify matters greatly to avoid "ISO" and use "exposure index" and "ISO setting", where they apply.

--
John

 

[Ali_baba]

However, there is a problem if you can demonstrate, that ISO 3200 is better than ISO 100 +1 EV in PP, because the 5D does not have real ISO 3200.

The pics showed no difference between ISO 1600 ev+1 and ISO 3200, which demonstrates that ISO 3200 on the 5D is simply ISO 1600 pushed a stop.

--
Ali Baba

 

[Ancient_Mariner]

This thread has been very informative for me - and yet, it will take me several times reading it to fully to grasp it.

For me, it goes against all photographic instinct that higher ISO could ultimately result in less noise. I actually experimented, I took two shots at same speed/aperture, one at ISO 800, appearing slightly underexposed, the other at ISO 1600 appearing slightly overexposed

At ACRaw, I noticed that the ISO 800 images did not have blown highlights (as expected) but no blocked shadows either. The ISO 1600 had a few blown highlights. I reduced brightness to achieve the same appearance as the ISO 800 image and cropped an area without blown highlights to check the noise.

The ISO 1600 sample was a bit noisier than the ISO 800, as I expected.

Any comments on that?

PS

The camera used was a Nikon D40. I don't have the samples right now, but I might post them later on

 

[richardplondon]

Every time you increase the exposure over indicated by the meter, you are USING a lower ISO!!

It would simplify matters greatly to avoid "ISO" and use "exposure index" and "ISO setting", where they apply.

quite - RP

 

[Ron Poelman]

Why wasn't this test conducted OVER "correct" exposure
(which is what expose to the right is)
and then pulled back to correct exposure level ?

In other words, instead of cranking up underexposed shots
which would be short on information ,
why not take overexposed ones which will have all/too much/blown
information and compare ?

Shift to the right is overexposure (to whatever degree)
so why not model with that ?

--

 

[richardplondon]

Actually, outside of Canon DSLRs, and some Nikons, digital cameras generally have read noise proportional to ISO.

OK, thanks for the correction.

One complicating factor for anyone seeking to test this technical comparability of different ISO settings, is that default processing is typically not linear. It appears the default is often to apply a contrast-boosting S-curve and to raise the blackpoint a little so as to crush the noisiest darks to black - whether in a RAW converter, or an in-camera JPG engine. This is AFAICT done for perceptual reasons, not scientific ones.

Using these non-neutral settings, a RAW shot underexposed 1 stop at ISO 100, and boosted 1 stop in conversion will look darker than a RAW shot taken with the same aperture and shutter settings at ISO 200. Correspondingly, the noise will look worse.

Repeat the experiment selecting a linear tonecurve and with "blacks" at zero, and the two images are indistinguishable, at least for my cameras (LX1 and K10d) and apart from the highlights. In other words, in a fair comparison it's equivalent, but real life is seldom fair unless you make it so.

This non-neutral processing would also probably overwhelm any ETTR-specific advantage, if that were tested for. IMO the success of ETTR involves being prepared not only to fiddle with camera settings and histograms, but also to get involved in the conversion tonecurve.

RP

 

[Barrie Davis]

Every time you increase the exposure over indicated by the meter, you are USING a lower ISO!!

It would simplify matters greatly to avoid "ISO" and use "exposure index" and "ISO setting", where they apply.

quite - RP

Forgive me asking, but does that "quite" mean the penny has actually dropped in regard of what I'm driving at with "... using a lower ISO?"
--
Regards,
Baz

I am 'Looking for Henry Lee ' (could be Lea, or even Leigh) and despite going 'Hey round the corner', and looking 'behind the bush', I have not yet found him. If he survives, Henry is in his mid-60s, British, and quite intellectual.

What is it all about? Well, something relating to a conversation we had in the pub 35 years ago has come to spectacular fruition, and I'd like him to know how right he was.

If you know somebody who could be this man, please put him in touch with me. Thank you.

 

[AnandaSim]

The ISO 1600 sample was a bit noisier than the ISO 800, as I expected.

Any comments on that?

I'm keeping a casual eye on this thread but really we have all kinds of photographers becoming expert technicians and scientists.

I'm from the school where you don't have to have a PhD to make a mobile phone call (mobile phone are incredibly complex technology) or click a shutter.

This to and fro about whether ISO increases noise or decreases noise is a secondary issue.

What I perceive is that the spirit of ETTR is you don't want low total data captured.

The histogram is a 2D chart. The chart is a statistic. Do you know about areas of of rectangles? Take the bar at 250. Say it is 10 units tall. So you have 10x250 = 2500 units. Now, multiply all the bars like that.

An histogram with ETTR will have a shift to the right - so there will be more total units. A histogram of similar tall and narrow shape could be centered or could be touching left toe to digital 0 on the X-Axis.

So compare the three tall and narrow histograms (we assume that all the histograms don't spread full width). The histogram with the most number of units will be the ETTR and will have the most amount of data.

Now, put on your visual "hat". You are now in front of the screen and are about to move your photo editor's sliders to adjust brightness, contrast, curves visually. The histogram with the least amount of units of data, if you manipulate it, will more likely solarise and more likely show more noise - i.e. there will not be enough units of data to spread like butter over bread .

If you don't have enough butter, it becomes too thin and you have bare bread and not butter. That's the noise.

--



Ananda
http://anandasim.mp

 

[Ehrik]

Has this read-noise improvement made its way to compact cameras, say, like my Panny LX3 I'm just learning to get the best from??

No. The LX3 has read noise almost perfectly proportional to ISO.

Which seems to be the case with all CCD cameras, right? Including DSLRs.

The behavior referred to is only true of Canon and Nikon DSLRs, maybe some others, but I haven't seen any evidence.

Aren't the DR curves at DxOmark useful for this purpose? In the ISO range where they lose less than one stop of DR per stop of (actual measured) ISO, raising the ISO will have a signal-to-read-noise benefit.

This would mean,

for 10Mp 4/3" cameras, up to ISO 400
for 12Mp 4/3" cameras, up to ISO 800 (*)
for Pentax 15Mp cameras, up to ISO 1600 (?)
for Sony A700, up to ISO 200 (maybe, hard to judge)

Then there could be reduced line noise on top of this, as you have pointed out elsewhere.

*) I've looked at blackframes from the G1/GH1/GF1 in IRIS and can confirm that the read noise doesn't go up proportional with ISO. Estimating the exact magnitude is hard due to black clipping. The GH1 has a lot less read noise than the other two. It uses another 12Mp sensor.

 

[richardplondon]

Forgive me asking, but does that "quite" mean the penny has actually dropped in regard of what I'm driving at with "... using a lower ISO?"

I only took issue with that particular form of words, because it would be false as a literal statement. " Equivalent to using a lower ISO setting " is perfectly fine, though, subject to some rather minor real-world effects where equivalence breaks down in practice.

We have largely been saying exactly the same things, but with some unfortunate conceptual dissonance. I'll happily tune that out in future.

RP

 

[Barrie Davis]

richardplondon wrote:
snip

We have largely been saying exactly the same things, but with some unfortunate conceptual dissonance. I'll happily tune that out in future.

Fine. :-|
--
Regards,
Baz

I am 'Looking for Henry Lee ' (could be Lea, or even Leigh) and despite going 'Hey round the corner', and looking 'behind the bush', I have not yet found him. If he survives, Henry is in his mid-60s, British, and quite intellectual.

What is it all about? Well, something relating to a conversation we had in the pub 35 years ago has come to spectacular fruition, and I'd like him to know how right he was.

If you know somebody who could be this man, please put him in touch with me. Thank you.

 

[John Sheehy]

Has this read-noise improvement made its way to compact cameras, say, like my Panny LX3 I'm just learning to get the best from??

No. The LX3 has read noise almost perfectly proportional to ISO.

Which seems to be the case with all CCD cameras, right? Including DSLRs.

The behavior referred to is only true of Canon and Nikon DSLRs, maybe some others, but I haven't seen any evidence.

Actually, now that I remember, the Pentax K20 had this behavior, too (the K10 didn't).

Aren't the DR curves at DxOmark useful for this purpose?

Yes, as long as the highlight headroom isn't changing for some ISO(s).

In the ISO range where they lose less than one stop of DR per stop of (actual measured) ISO, raising the ISO will have a signal-to-read-noise benefit.

This would mean,

for 10Mp 4/3" cameras, up to ISO 400
for 12Mp 4/3" cameras, up to ISO 800 (*)
for Pentax 15Mp cameras, up to ISO 1600 (?)
for Sony A700, up to ISO 200 (maybe, hard to judge)

Then there could be reduced line noise on top of this, as you have pointed out elsewhere.

*) I've looked at blackframes from the G1/GH1/GF1 in IRIS and can confirm that the read noise doesn't go up proportional with ISO. Estimating the exact magnitude is hard due to black clipping.

You can do a histogram analysis of the blackframe to get a rough estimate of where black actually is.

Let z = number of zeros. Let y = number of non-zeros. Let p(potential real zeros) = z-y. Does p seem like the peak in a natural, gaussian progression, going

+3 -> +2 -> +1 -> p? If so, black is probably clipped in the right place. If not, slide the histogram and try again.

I have found that if the blackframe is clipped exactly where it should be, and there is enough bit depth so that there is little quantization (a nice half-bell forms), then the real standard deviation is about 1.63x the measured one.

The GH1 has a lot less read noise than the other two. It uses another 12Mp sensor.

What was the std dev as measured, and what is the bit-depth?

--
John

 

[John Sheehy]

richardplondon wrote:

Forgive me asking, but does that "quite" mean the penny has actually dropped in regard of what I'm driving at with "... using a lower ISO?"

If I have my camera set to ISO 100 and 0EC, and this results with a normal histogram or exposure, and then I change to ISO 200 and +1 EC, I have increased the "ISO" in the most commonly accepted usage. I can clarify by saying that the EI (exposure index) is still 100, but I did change the camera's ISO setting.

Your "point" could be stated much more clearly as "shoot with as high as an absolute exposure as possible", or "shoot with the lowest EI possible". However, for the least noise, one should also add that the highest "ISO setting" on the camera that doesn't clip wanted highlight detail gives the least noise (up to the highest analog gain, or just short of it).

--
John

 

[John Sheehy]

We have largely been saying exactly the same things, but with some unfortunate conceptual dissonance. I'll happily tune that out in future.

Barrie must word things in such a way that "low" ISO comes out on top.

See if you can get him to admit (or even believe) that the highest ISO setting that doesn't clip the desired highlights gives either the least noise or the same noise on most cameras - he still may be far from saying that.

--
John

 

[John Sheehy]

This thread has been very informative for me - and yet, it will take me several times reading it to fully to grasp it.

For me, it goes against all photographic instinct that higher ISO could ultimately result in less noise. I actually experimented, I took two shots at same speed/aperture, one at ISO 800, appearing slightly underexposed, the other at ISO 1600 appearing slightly overexposed

At ACRaw, I noticed that the ISO 800 images did not have blown highlights (as expected) but no blocked shadows either. The ISO 1600 had a few blown highlights. I reduced brightness to achieve the same appearance as the ISO 800 image and cropped an area without blown highlights to check the noise.

The ISO 1600 sample was a bit noisier than the ISO 800, as I expected.

Any comments on that?

PS

The camera used was a Nikon D40. I don't have the samples right now, but I might post them later on

The Nikon D40 was one of the first Nikons to have read noise non-proportional to ISO, dropping at higher ISOs. You or your converter are probably doing something non-equal.

Most camera either will have less read noise (relative to absolute signal) at the higher ISO, or the same. The only camera I've seen where higher ISOs are absolutely noisier is the Panasonic FZ50 I have, and only by a small amount.

--
John