Expose to the right (ETTR) or underexpose?

[sblitz]

Advertisement (gone after registration)

So if the situation you are shooting is irrelevant to shutter speed and aperture you are better off Ettr at the base iso and raise the exposure in pp?

[Lindolfi]

similar to these NEX files?

 

No, the banding I referred to is in the image, not in the histogram. The histograms of the 3 stops pushed 160 ISO images are not combed like those of the NEX.

[Lindolfi]

Could you pick up the alleged NR that Leica applies @ high ISO? ( maybe only at 2500?)

 

When a 4 stop push is applied to a 160 ISO exposure and compared to 2500 ISO in 6500 Kelvin light, the luminance noise is on average equal, but not the same in every color patch of the MacBeth card. Sometimes one is better, sometimes the other. The shapes of the luminance histograms is different (pushed 4 stop 160, looks Gaussian, so bell shaped and 2500 ISO looks more like the Monument Valley mountains, so narrow middle part with wider outskirts).

 

However the color noise is a bit higher in the 4 stops 160 ISO exposure than in the 2500 ISO exposure. Have not measured it, but in the dark gray patch of the MacBeth card you see vague magenta blotches of about 50 pixels diameter that are not present in the 2500 ISO exposure. This difference is not visible in the earlier 3 stop pushed 160 ISO versus1250 ISO comparison I did.

[douglasf13]

Would a Nikon camera behave similarly to the 5DII?

 

It depends on the Nikon camera. The D700, D3, D4 and others will behave like the 5Dii in this regard. The Nikon cameras with Sony sensors, the D7000, D3x, D800, etc., will perform more like the M9. It all depends on the design of the sensor and ADCs.

 

A good indicator is by looking a the DR measurements at DXO Mark. If the DR graph is linear, the camera will likely be fine with pushing exposure in post. If not, it's probably better to do it in camera. In fact, for the cameras with non-linear DR, the ISO where the DR does become linear is likely where the camera can be pushed from that point. In other words, but I'd bet that you could push the exposure in post once you get to ISO 1600 with the 5Dii.

[douglasf13]

So if the situation you are shooting is irrelevant to shutter speed and aperture you are better off Ettr at the base iso and raise the exposure in pp?

 

I'm not sure I understand the question. The overall point is that shutter speed and aperture are really the ONLY things relevant, and ISO is less so. ETTR is about getting as much light to the sensor as possible, and it is independent from ISO. ISOs above base ISO are simply camera gain added to the signal AFTER the exposure, rather than being a part of exposure itself (although they do make it easier to calculate exposure.)

[douglasf13]

Could you pick up the alleged NR that Leica applies @ high ISO? ( maybe only at 2500?)

 

According to DXO Mark, the M9 applies NR to raws at all ISOs from 320 and up, and "This noise filtering is very slight but constant along the whole dynamic. It increases slightly with ISO."

[douglasf13]

Advertisement (gone after registration)

Having those higher registers of the histogram more populated with information, makes a LOT more sense to me.

 

I don't think anyone is arguing that more light hitting the sensor improves SNR. That is certainly so. The point is that ETTR doesn't make sense above base ISO with any camera, and that is a commonly misunderstood point about the technique.

[wlaidlaw]

It will be interesting to see when the M-240 comes out, if the improved Maestro processor has a better signal to noise ratio than the current processor of the M9. I assume that it is inadequacy of the in camera ISO amplification process that is producing an inferior result, in comparison to boosting the exposure in the RAW processing software. Has anyone got an S2 that they could run the tests on. Again of course, the type of signal that comes from the different read process of a CMOS sensor in the M-240 may lend itself to quiet in camera amplification, more than the signal emanating from the CCD of the S2.

 

Wilson

[Jaybob]

This is where I'm hanging my hat...

 

There is more data captured the farther to right it is captured – much more actually. So if you have a choice you want to move the tonal values as far to the right as possible without blowing the highlights.

 

If a sensor resolves, say, 16384 values (14 bits) and you divide the histogram into 14 zones of one EV each, there are 8192 different values resolved within the rightmost zone, 4096 in the next, then 2048, 1024, 512, and so on.

 

 

Even with a compressed NEX file, at base ISO, there is so much more tweakable information in this histogram, then if I intentionally underexposed it.

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

[Lindolfi]

Jaybob, the argument that ETTR is better than ETTL (Expose to the Left) on its own is beyond dispute. However, there is a situation in which it does not necessarily apply: in the case where you are unable or unwilling to increase exposure of the sensor. In that case ETTR plus increase of ISO produces similar results as EWIP (Expose What Is Possible) at base ISO and pushing in post processing. It is up to you what you do, but the ETTR argument on its own is not sufficient to cover all options.

[Jaybob]

...ETTR doesn't make sense above base ISO with any camera, and that is a commonly misunderstood point about the technique.

 

I went back to my original post and looked again at the first file I uploaded. Shooting in shade at 800 Aperture priority, my groom walked out into full sunshine. It's a real world situation. That's where the method makes sense, as far as adjusting all that upper histogram awesomeness.

 

Honestly, in the real world, you'd all think I was the greatest guy you'd ever met. Somehow, in here, I seem like a douchebag.

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

Edited October 27, 2012 by Jaybob

[Lindolfi]

1/4000 sec. at 800 ISO? You could have exposed much more sensibly here! You had so much light that the whole argument of needing to do anything else than ETTR does not apply. But you could have less noise by going to base ISO and exposing longer.

 

(Given the histogram and the sunlight this must have been about f/5.6 at 800 ISO and 1/4000, not f/1, probably this got in the exif due to the lack of communication between lens and body)

 

PS Could not help noticing you added a remark in your last post about what kind of person you seem to be. That is not the issue at all. The fact that you love photography already hints at a likeable person. However, this is about discovery and learning from each other, not who we seem to be.

Edited October 27, 2012 by Lindolfi

[Jaybob]

YES Lot of light agreed, but again, this was a real world, real time, and very fluid situation.

 

Going from full shade at f4 or f5.6 where I was evenly lit at about 125th of a second to a perfect exposure in full blast sun, white jacket, dark skin. I barely had time to focus.

[Lindolfi]

OK Jaybob, we have all been there: no time to change ISO. But in these conditions with this ISO it is not the best example to use as an argument in a thread on optimal exposure. But before you come up with another example: I fully understand your arguments and as I said, I think they are valid, but do not cover all the options in all conditions.

Edited October 27, 2012 by Lindolfi

[mjh]

It will be interesting to see when the M-240 comes out, if the improved Maestro processor has a better signal to noise ratio than the current processor of the M9.

A digital processor doesn’t have a signal to noise ratio. There is shot noise (the noise inherent in the light itself), various sources of noise within the sensor, read-out noise, and quantisation noise, but once the camera can deal with digital data, no additional noise gets added.

 

I assume that it is inadequacy of the in camera ISO amplification process that is producing an inferior result, in comparison to boosting the exposure in the RAW processing software.

Analogue amplification adds noise that a simple digital multiplication in the processor would not, but on the other hand an amplification can reduce quantisation noise (and increase the number of tonal values resolved). For every camera there is an ISO setting where the total signal to noise ratio ceases to benefit from additional analogue amplification; above that point analogue amplification before quantisation can be replaced by digital multiplication after quantisation. And if one doesn’t employ a simple-minded multiplication but primarily adjusts the shadows and midtones, one can even increase ISO without sacrificing dynamic range.

 

Has anyone got an S2 that they could run the tests on. Again of course, the type of signal that comes from the different read process of a CMOS sensor in the M-240 may lend itself to quiet in camera amplification, more than the signal emanating from the CCD of the S2.

It’s the analogue circuits that matter and these have nothing in common between the S and the new M.

Edited October 27, 2012 by mjh

[Lindolfi]

Jaybob iimagine, you have an 75/1.4 on your M9 and you are asked to photograph a ballet. You get an optimal seat front side of the stage. You measure the light and you realise that f/1.4 is needed and possible. Exposure time is 1/30th at ISO 800 and you know that that is also possible. You can not expose longer, because the dancers will show nothing sharp, while at least the head should be sharp for what you have visualised.

 

Now you have a choice: use f/1.4 at 1/30 at 800 ISO, or use f/1.4 at 1/30 at 160 ISO. Both will give you good results. But given the highlights from the stage lights on legs of dancers close by at some moments during the choreography, you have a bigger chance to get good images with underexposure at 160 ISO.

 

Hope this makes clear what I mean with those conditions. By the way, this was one of my real world situations.

 

PS oops! I see you have removed your last post in which you asked a question, to which this was my answer.

Edited October 27, 2012 by Lindolfi

[Jaybob]

Sorry, but if it's that dark, I reach in the bag and use a D3 set on 3200, switching between the 50 and 85 1.4 or the 105 f2. That gets me to a wide open 125th or a 160th or a 200th with fixable noise later. There's a limit as to what's hypothetical and fun, and paid jobs providing usable frames to clients are neither. A 30th, to me, will yeild far too many "not quites". Just my opinion, the M9 is the wrong tool for that job, at least at the beginning. After getting what the check writers need written to the card on the D3, I would switch to the M9 and look for lulls in the movement (or exploit the movement) and finish the job as a photographer. Does the 75 1.4 come with this deal, cause I need one of those...

 

ETTR specifies expose for highlights and don't blow them out. That's what I'd do regardless of camera or ISO. If anything in my histogram is hot, that's a faster shutter speed I can use.

 

Then I'd find the lighting director and ask him what his qualifications were and expose him for the fraud he or she really is. That seems REALLY dark for spot lit (or even just LIT) performers.

Edited October 27, 2012 by Jaybob

[Lindolfi]

I'm sorry too, for not being able to make my point clear with this example. In your last post there are too many points raised outside the main subject of this thread about photography to be able to react to them in this thread without breaking up the main line of the conversation, so I leave it like it is for now.

[wlaidlaw]

It’s the analogue circuits that matter and these have nothing in common between the S and the new M.

 

Michael,

 

Does the analogue amplification (ISO setting) occur therefore within the sensor electronics prior to the signal being digitised and is the basic A-D conversion done by the sensor or the processor? I had assumed wrongly that this amplification was done by the processor and then the processor handled the A-D conversion. If it is being done at the sensor level, I accept that any result from the S2 with its CCD sensor would be wholly irrelevant to the M-240.

 

Wilson

[Advertisement]

Hallo,

Look here: Leica M

[Jaybob]

I'm sorry too, for not being able to make my point clear with this example. In your last post there are too many points raised outside the main subject of this thread about photography to be able to react to them in this thread without breaking up the main line of the conversation, so I leave it like it is for now.

 

I think the kids say "I know, right?"

 

I was attempting to provide examples of how and when I try to use ETTR opposed to the examples that were posted here before.

Edited October 27, 2012 by Jaybob