selmslie Loc: Fernandina Beach, FL, USA

All raw converters are non-linear or they would not work.

The voltages read from the sensor double with each one-stop increase in exposure. A 14-bit raw file can represent 16,384 distinct tonal values but there are 8,192 values in the brightest zone, 4,096 in the next zone down, etc., until at about 10 stops down there are only 8 gradations. This is the foundation of the ETTR rationale.

But we see tonal values in an arithmetic rather than a geometric fashion. The geometric progression in the raw file is mapped in a raw converter to an arithmetic progression in which each zone ends up with the same number of values. Converting from a geometric progression to an arithmetic progression is always a non-linear process.

When you finally get down to an 8-bit JPEG (256 possible values), you end up about 25 gradations within each zone of ten-zone image. A 16-bit TIFF with a total of 65.536 possible values end up with about 6,500 values in each zone.

For a JPEG of a raw file for a 10-zone subject, Zone X needs to squeeze the 8,192 raw values into about 25 JPEG values. Zone I needs to extrapolate from the 8 raw values to get the 25 JPEG values.

If you have a ten-zone subject you are pretty much out of luck whether or not you use ETTR. You are still going to end up with poor shadow detail and/or the risk of blown highlights. It is only subjects with fewer than a 10-zone range that might benefit from ETTR. But for these you are more concerned with minimizing noise than with available tonal values. And the best way to avoid noise is to lower the ISO until you reach the camera's native ISO, about 100 or 200.

Uuglypher Loc: South Dakota (East River)

Hi, Scotty,

So once again , you are the "hired gun" for the "take my word for it, EBTR just doesn't work" crowd. Here again to try dazzeling with footwork?

you said ( again):
"All raw converters are non-linear or they would not work."

The point, Scotty, is that although your raw converter of choice produces a TIFF file without tonal or color normalization which can then be tonally and chromatically modified by linear processing in a totally non-destructive manner, if that Tiff file is moved into the non-linear environment of Photoshop for tonal manipulation, then as soon as you compress ... and then expand the file's data you'll find you've lost some of the image data ...leaving characteristic gaps and spikes in the histogram...which does not happen, thanks to linear procesing, in the Raw converter.

Allso, Scotty, your concerns about the tone/zone/value /stop scale of the zone system being difficult to encompass within the JPEG dynamic range of our cameras ( whether they initially capture 12 or 14 bit-depth data) is , indeed , a bit out of date, with many camera's sensors now providing 11, 12, or 13 Stops/E.V.s....and that's just for JPEGS.

When you carefully determine the extra RAW-accessible dynamic range (ERADR) beyond the right limit of the JPEG-adjusted histogram and its associated clipping warning, you may find an additional 2/3 stop to more than two full stops to utilize for RAW capture. And when doing so the "darker" end of the exposure's " light pile" is captured well to the right of the dreaded noise zone where low signal-noise ratios are prevalent due to low exposure. Underexposure results in fewer photons collected by each photosite (pixel) and the wages of noise generated as the square root of the number of photons captured becomes most evident in data collected there. This is in contradistinction to the result of greater exposure of the "dark data" captured further "to the right" where the square root of the greater number of photons captured by each photosite translates as proportionately less noise in the darker regions of the image...and practically no noise in the brighter regions. That's the reason that EBTR does, in fact, result in less noise, even in the face of use of ISOs as high as 800 and 1600. It is also well known among its routine practitioners that ETTR provides practically no benefit to JPEG capture.
In spite of your first experience with ETTR having been " complicated" and not worth the trouble as you have reported to me in another conversation, you really ought give it...and then EBTR ... another try. Put down the pencil and slide rule and quit sounding like the aeronautical engineers who persist in denying that the bumblebee can fly ( on paper).
Buy or rent a decent modern DSLR with a DR of 11, 12, or 13 Stops , test its ERADR, and then use it properly in ETTR and then in EBTR. I'd love to see your reaction when you see the results in practice as opposed to what they ought be according to "theory".

Cheers,
Dave

selmslie Loc: Fernandina Beach, FL, USA

I never said it does not work, just that it only works in a limited number of situations and that is becoming less important than it was a decade ago.

Your subject, at least those portions of the scene about which you care, must have a limited dynamic range, perhaps eight stops or less, to give you some elbow room to over- or under-expose by a stop without losing significant texture or detail. In this case, overexposing one stop might shift portions of you scene that would normally record in Zone II through Zone IX up to Zone III through Zone X. You would then lower these values during conversion or in post processing to restore the original distribution of tones. The rational for ETTR is that this would avoid some noise in the lower zones and retain the subtle graduation of values. We are all on the same page here.

A digital raw file records a geometric progression of tones which must be converted to an arithmetic progression in either a JPEG or a TIFF file. This conversion is non-linear and it involves interpolation at the high end extrapolation at the low end of the scale. It might also involves noise, which is more visible at the low end than at the high end.

There has been some significant progress in digital imaging since 2003 when ETTR was first proposed. The problems for which ETTR was developed have not gone away; they have just become less significant. Small format DSLRs (including full-frame 24x36mm) now provide 14-bit and medium format can have 16-bit raw files. Sensors are larger and less susceptible to noise.


I already have a decent modern DSLR, a D610. I have never run into an issue with noise or tonal graduations, even in sunny Florida and even with a lesser D7000, mainly because I keep my ISO at 800 or less.

And, as you know, when I face a challenging scenario I just use medium format film, but that’s a whole separate conversation.

Uuglypher Loc: South Dakota (East River)

Well, Scotty, ( and those for whom you attempt to carry the water...)

You mention that: "The problems for which ETTR was developed have not gone away; they have just become less significant."
Agreed, and I'm sure that trend will continue, but as long as the problems of image data quality as regards rendition of detail relative to bit depth and detail degradation by noise remain to any degree "significant" , like it or not, Guys, EBTR for maximal benefits of RAW capture will remain the remedy of choice!

And Ah hah.., Scotty,
now you admit to keeping ISO to "800 or less" (Sounds like you actually are now giving it a try at higher ISOs than your initially recommended " native" ISO !!!! In fact, if you'll actually try it, with EBTR, ISOs considerably higher than 800 will also benefit greatly in terms of noise reduction.)

I'm glad to be sensing a slow, gradulal (albeit reluctant) acceptance of the facts that explain why those of us who have routinely used EBTR for years continue to do so.

Such bifuscating footwork conflating the detail-rendering abilities of 8-bit, 12-bit, and 14-bit image data files reflects, at best, downright ignorance of what's attainable in practice, or at worst, simply dishonest. It is an incontrovertible fact that there is no contest between the detail rendition in a 16x20 (or even much larger) print from an 8-bit JPEG file on the one hand and the same sized print from either a 12-bit or 14-bit TIFF file prepared from a RAW capture ....regardless of their having being worked in a 16- bit environment ...on the other hand.

The relative numbers of potential cusps of detail provided by 14 bits versus 8 bits clearly tell the tail! (see the pie chart graphic)

Cheers,
Dave in SD

Searcher Loc: Kent, England

Selmslie and Uuglypher

I am not joining in your interesting but bewildering discussion, but I would like to ask a question -

Whilst ETTR means Expose to the Right but do not overexpose, does EBTR mean you actually do deliberately over expose the highlights? I am sure I have got that wrong, but would welcome an answer in ordinary English please.

selmslie Loc: Fernandina Beach, FL, USA

I actually keep my ISO low because I don't need to take photos in low light. In fact, high ISO situations that have a noise problem get the same benefit from simply lowering the ISO. After all, ETTR makes you open the aperture and/or lower the shutter speed to let in more light.

For example, If you shoot at ISO 1600 and, because of the noise, your ETTR exposure is 1-1/3 stops higher, you get the same amount of light on the sensor and the same amount of noise as if you had shot it normally at ISO 640 in the first place.

The benefit of ETTR is that you skew the tonalities recorded in the raw file to provide more steps, but with ETTR you don't get to preview your results in the JPEG in the camera's display.

You can find a clear description of the pros and cons of ETTR and other exposure methods at http://www.cambridgeincolour.com/tutorials/digital-exposure-techniques.htm

Incidentally, you might want to re-draw your pie chart. The slice for the JPEG should occupy less than a degree and a half (360/256). It looks like the red wedge is closer to 5 degrees.

Scotty

selmslie Loc: Fernandina Beach, FL, USA

Overexpose is relative. You overexpose so far as the JPEG is concerned but you do not (you hope) use up the raw file's additional latitude.

Uuglypher Loc: South Dakota (East River)

Hi, Searcher,
it's a logical question, but in the context of exposing RAW data EBTR is most definitely NOT overexposure! When we are exposing RAW DATA we are just using the JPEG-adjusted histogram and the so-called ETTR exposure as the "jumping off point" for adding exposure of the extra RAW-Accessible Dynamic Range (ERADR) that we have determined our particular camera to have "in reserve" off to the right of the JPEG-adjusted histogram. Utilizing the ERADR is necessary to adequately expose RAW data to assure the maximum of image data quality.
If we were, in fact, to be exposing for a JPEG file, EBTR would, indeed, be overexposure!

Please let me know if that helped. There has been a lot of obfuscatory and unnecessary data presented here in the interest of "dazzel'em with footwork" to distract (falsely impress?) the unititiated from the demonstrable benefits of EBTR.

Dave in SD

Uuglypher Loc: South Dakota (East River)

"I actually keep my ISO low because I don't need to take photos in low light."

And that, Scotty, points out a great advantage of EBTR; it permits use of higher ISOs permitting exposure under low light conditions without the severe increase of noise otherwise experienced!

As for the pie chart, Scotty, it is perfectly accurate. It illustrates the 8-bit JPEG file's small proportion (256 =1/64th) of the 16,354 cusps of 14-bit detail offered by RAW capture. (5.625° is, in fact 1/64 of 360°). i actually think that for eyeballing 1/64 th of the circle's cicumferance I did pretty well....certainly well enough to make that inconveniently embarassing point, eh?

Dave

Searcher Loc: Kent, England

Selmslie and Uuglypher

Thank you both, I had not realised that ETTR and ETBR both related to histograms, I have always used the "right" as being the most exposure possible without blowing out the highlights. My mind is still in the world of film, no histograms to look at.

selmslie Loc: Fernandina Beach, FL, USA

You still do not explain how that is different from just using a lower ISO.

Yes, I was thinking of a TIFF file.

Uuglypher Loc: South Dakota (East River)

The point, Scotty, is that regardless of the ISO, use of EBTR reduces whatever level of noise would otherwise occur because it pulls the left ( dark) end of the light pile away from the left end of the JPEG-adjusted histogram where the greatest amount of noise always lurks!

Dave

selmslie Loc: Fernandina Beach, FL, USA

What reduces the visible noise is the additional exposure from the larger aperture and/or longer shutter speed. It's the S/N (signal to noise) that changes. This has nothing to do with ETTR, EBTR or how you meter it whether via the histogram, incident or spot meter.

The only effect you achieve from setting the ISO higher and then overriding the recommended exposure is that you shift the encoded bits in the raw file. This makes more use of the higher values in the raw file to provide smoother transitions in the tonality, but it only matters if you need better low zone tonality smoothness. The middle and upper zones are already well represented in the 14-bit raw file. Any improvement to the tonality in the middle zones will be impossible to see if you convert the 14-bit raw file to a 16-bit TIFF.

Another danger you risk is that not all channels (red, green and blue) might get clipped at the same time. Your camera may not tell you this in the histogram but you should find out later in your raw editor. The safest thing to do is to not trust your histogram or meters. Simply bracket your shots and select the best exposure when you do the raw conversion.

When you subsequently work with the 16-bit TIFF, all of the information will have been evenly distributed to all of the zones from black to white.

Uuglypher Loc: South Dakota (East River)

Scotty stated:!"What reduces the visible noise is the additional exposure from the larger aperture and/or longer shutter speed. It's the S/N (signal to noise) that changes. This has nothing to do with ETTR, EBTR or how you meter it whether via the histogram, incident or spot meter."

Indeed, Scotty, you are right that what reduces noise is the change to a greater S:N.   It changes precisely because you use the extra RAW-Accessible Dynamic Range (ERADR) to capture RAW data via the process of EBTR.  It has EVERYTHING to do with utilization of the ERADR by EBTR !

Xxxxxxxxxxx

Scotty said: "The only effect you achieve from setting the ISO higher and then overriding the recommended exposure ...."

This leading premise is false!   You set a higher ISO precisely to be able to use a faster shutter or a smaller aperture and, by utilizing the ERADR for EBTR, you incorporate less noise in your image even than the amount you would have introduced with a lower ISO had you stayed within the "recommended exposure for JPEG" .  In using an optimal exposure for RAW data you are NOT " overiding the recommended exposure!

The recommended exposure for a JPEG file is very different from that necessary to assure proper exposure of RAW data in order to gain the most benefits inherent in RAW image data capture.  Optimal exposure for,JPEG is not equal to optimal exposure for RAW data.

Xxxxxxxxxx

Scotty said "Another danger you risk is that not all channels (red, green and blue) might get clipped at the same time. Your camera may not tell you this in the histogram but you should find out later in your raw editor. The safest thing to do is to not trust your histogram or meters. Simply bracket your shots and select the best exposure when you do the raw conversion.

That might be so...if so, it hasn't reared its head as yet ( in 5 years) If it does, I'll surely be grateful for this advice.

The film photographer is well within his comfort zone while metering and exposing for JPEG image files, and even for capture of RAW data within the constraints of the JPEG-adjusted histogram and the displayed JPEG thumbnail image provided by the camera.

Unfortunately, here is where the traditional photosensitive emulsion photographer continues to attempt to establish equivalences between film photography and RAW image data just as he has with JPEG image file capture. Those equivalences with RAW capture simply do not exist.

Best regards,
Dave

selmslie Loc: Fernandina Beach, FL, USA

Assume you have your ISO set to 1600 and you capture the scene as the meter recommends - 1/250 and f/11. According to the histogram you are almost but not quite clipping the highlights and shadows. You are pretty sure that you will have noise at the low end but your camera has about 1 stop of extra latitude at the high end for the raw file. So you set your exposure for an additional stop as either 1/125 at f/11 or 1/250 at f/8. In other words, you have overridden the "normal" exposure recommended for ISO 1600 by +1 stop in order to expose to the right.

Or you could drop your ISO to 800. The meter now recommends a setting of 1/125 at f/11 or 1/250 at f/8, exactly the same setting that you ended up using with ETTR.

Both images will end up with the same S/N ratios throughout the scene. So it turns out that you do not use a higher shutter speed or smaller aperture with the ISO 1600 setting.

The only difference between the two images is how the information gets recorded in the raw file. The ETTR file might encode it as Zones III through X and the lowered ISO version as Zones II through IX.

Both images might end up looking exactly the same after adjusting the exposure during the raw conversion, but you will not know this until you convert them.

There is still the possibility that you could have posterization with the ISO 800 version that you might have avoided with the ISO 1600/ETTR. But have you ever seen posterization in the shadows or low zones? So far you have not mentioned it. I have never seen it. Neither am I worried about noise since I just use low ISO settings with my 14-bit raw files.

Again I refer you to http://www.cambridgeincolour.com/tutorials/digital-exposure-techniques.htm for a clear presentation of the pros and cons of different ways to set exposure, starting with ETTR.