Exposure Compensation or higher ISO??...

The advice in this thread is very confusing.

The answer to the question is not.

Yes, you can do this with digital. Just shoot in RAW, use a large aperture and underexpose one to two stops to get the fastest shutter speed with the available light.

Take the RAW file and bring it back to life on your PC.

I have heard comments like a well exposed image is better and yes, it probably is. But if you are somewhere without a tripod and the shutter speed is too low at ISO 100, your choices are limited. This technique gives better results than upping the ISO to maximum.

See my full article and example images here:

http://www.steve-pearce-photography.com/10dlow​lighttechnique/

Hi,
You can also look at it from the digital domain.
A properly exposed picture will have values from 0 (black) to 254 (peak white)
This is your dynamic range or contrast ratio.

When you underexpose you will loose resolution in the top end, meaning the picture will look very dull, when you shoot the right something else happens, by extending the black down with curves you can get a more dynamic looking picture because somehow the human brain will forgive lack in black detail better than lack in white/middle gray detail.

By getting a higer ISO and proparly exposing with the modern machines the noise is acceptable, much better in my opinion than for example underexposing a 800ISO shot.

Greetings,
Frank

ref:

"Well, did it not occur to you that shooting at ISO 800 and then overexposing during post processing is doing exactly that, increasing the image gain by digital amplifcation even though you may not think of it in that light when you are using Photoshop"

My thoughts are that PS CS will do the post processing better than digital amplification because there is no electronic noise component involved - i.e. its a much cleaner image afterwards. Am I wrong ?

Alan

slin100 wrote:
Sure. UseNet. http://groups-beta.google.com …earch&&d#c3bdeb​86d8a09f73

Before you go discounting the claim, read John Sheehy's post. John Sheehy is a pretty sharp guy. I see no reason for him to and no indication that he has lied or drawn the wrong conclusions. I believe he verified his findings by directly examining the RAW data using DCRAW, an open-source RAW decoder. In another post of his, he claims that the values in an ISO 1600 RAW file were all even, confirming his hypothesis that the A/D output was doubled.

Now, can you tell me why you disagree with me?

Well, he seems to contradict himself in his description of what happens.

Rephrased, he appears to be saying that:

1. At ISO 100-800 the signal's run through a linear amp.
   
2. At ISO 1600, the 1-stop underexposed signal's run through a linear amp at the ISO 800 setting and then doubled
   
3. At ISO 3200, the 1-stop underexposed signal's run through a linear amp at the ISO 1600 setting and then doubled.
   

You bring up a good, valid point. I believe your interpretation is correct and mine is wrong. I found another posting from John that backs up your interpretation.

I don't know. Perhaps the amp starts to behave non-linearly when pushed all the way to ISO 1600. Unfortunately, I have no evidence to back this conjecture. That might explain why ISO 3200 needs to be enabled.

He did make a distinction. The difference is where the gain is being applied: in the analog stage or the digital stage. His point is that it's mostly analog, but that there's 1 stop of digital amplification going on at ISO 1600 and ISO 3200.

Yes, he says that ISO 1600 and ISO 3200 RAW data values are only even, which corresponds to a 1-bit loss of precision.

spearce6 wrote:
See my full article and example images here:

http://www.steve-pearce-photography.com/10dlow​lighttechnique/

Steve,

I looked at your article. Your advice is similar to mine except that you recommend to start using negative EC at lower ISOs, even ISO 100 with EC -1 instead of ISO 200.

It's difficult to draw conclusions from your images. You don't provide 100% crops of the same image for both the ISO 400 with EC -2 shot and the ISO 1600 shot. The ISO 400 crop is from a dim portion of the image but the ISO 1600 shot is of the bright, central part of the dome. It's much harder to perceive noise in dim areas than bright areas.

In my experience, the noise from the 10D is quite low up to ISO 400. I believe that resorting to negative EC below that ISO is not necessary.

But how does the hypothetical "digital stage" differ from a two-stage linear amp dealing with analog data inputs all the way? And if 1600 off the linear amp is showing some problems, why use it rather than 800 quadrupled? Conversely, why not use the 1600 off the linear amp instead of 800 doubled for your 1600 value if the 1600 value's good enough to use as the basis for 3200 calculations?

slin100 wrote:
Furthermore, the kids are awake and I have to serve up breakfast!

lol.......thanks for some comic relief

Slin,

Maybe my article is a little unclear, sorry.
Both the dome 100% images are at ISO400, underexposed two stops and then taken back from RAW in C1.

Only the last two images on the page are at ISO1600.

The 10D noise is low up to 400 I agree, but often 400 is not even enough to get a handheld shutter speed - thus in the Basilica I used ISO400 with -2 stops to give an effective 1600. The last two images in the article show a comparison of a shot at ISO 1600 so that you can see the merits of 400 -2 stops.

I don't suggest resorting to negative EC below ISO400 - I have shown a table of shutter speeds and ISO's for comparison, but there is no reason not to take a negative EC shot with a low ISO. I have not made comparison images of noise at say ISO200-1 vs ISO400-2, but there is no reason not to try. There might be a slight benefit.

I don't agree that it's much harder to perceive noise in dim areas - quite the opposite. I think noise is always more noticable in dark areas of images.

The difference is that a digital multiplier stage magnifies the quantization error produced by the A/D converter. If the error coming out of the A/D is x, then doubling the values will increase the error to 2x. Quadrupling the quantization error will probably exhibit quite noticeable posterization effects.

That I cannot answer because I don't know what went through the designer's minds. That ISO 3200 is normally unavailable unless ISO Expansion is enabled in the Menu suggests the possibility that the designers may have felt uncomfortable with using a linearly amplified 1600.

Did you consciously choose ISO 400 with EC -2. Why not ISO 800 with EC -1? In your experience does the former exhibit lower noise?

In the audio world, it's quite important to match the sound levels when doing A-B comparisons. Some say that even a 1/2 decibel difference in volume can tilt the balance in favor of the louder system. Similarly, it's important to match the exposure levels between two digital images to provide a fair comparison. It's virtually impossible, IMO, to draw any conclusions from your examples because the exposure levels are different and the two images are different, as well! Of course, I can see the noise in the ISO 1600 100% but without an identical crop at ISO 400 with EC -2, well ...

You have demonstrated, OTOH, what can be achieved by aggressively pushing ISO in the digital domain.

I think I didn't articulate myself well. I meant to say that it's easier to perceive noise in the dark areas when the levels are brought up to accentuate the differences.

This evening I conducted a little test of my own. I made two test exposures. One was shot at ISO 400 with -2 EC and then converted to TIFF with +2 exposure increase. The other was made at ISO 1600 with 0 EC. The conversion was done (using EVU) "as shot" except for the +2 stop exposure increase for the ISO 400 shot to compensate for the -2 EC. Also, I adjusted the WB on both to correct WB. The shots were then loaded into PS so that I could resize them and make the 100% crop images. No editing or sharpening was done in PS.

Here are the EXIF info from the two shots:

These are the resized images of the two test shots:

HTTP response: 404 | MIME changed to 'text/html' | Byte size: ZERO

HTTP response: 404 | MIME changed to 'text/html' | Byte size: ZERO

HTTP response: 404 | MIME changed to 'text/html' | Byte size: ZERO

HTTP response: 404 | MIME changed to 'text/html' | Byte size: ZERO

HTTP response: 404 | MIME changed to 'text/html' | Byte size: ZERO

HTTP response: 404 | MIME changed to 'text/html' | Byte size: ZERO

Nice PacAce!! I do notice that the ISO400 pushed image tends to be a tad bit on the cool side as well. When being pushed, the noise seems to be blue, where the noise at ISO1600 seems to be a little green.BTW, which body did you shoot that with?... the 10D or 20D? The only reason I ask is because those images of yours seem to be a heck of a lot cleaner than mine at ISO 1600.

I wouldn't put too much stock in the overall color cast of the images since the WB correction may not have been exactly the same for both images. I tried to get them as close as possible but doing that manually is a little tough. However, you are correct in your assumption that there will be a slight color shift when you adjust the exposure by a stop or more.

I used the 10D for my test since that was the camera being discussed, if I recall correctly.

What I have learned about the high ISO settings, on the 10D at least, is that if you correctly expose the images, you will get the minimum of noise generated. Also, the darker areas such as shadows will show more noise than the brighter areas in the highlights. This is the reason shooting to the right of the histogram without blowing the highlights is so much endorsed. Comparing two images shot at, say, ISO 1600, the darker of the two will reveal more apparent noise than the other.