14n histograms, ERI and DNR

The 14n has some interesting and unusual exposure features. It implements a capability called ERI (Extended Range Imaging) and offers two dynamic range curves for JPEGS, wide (called portrait) and normal (called product). It's histogram is unique in that it added two features. The first is a ERI zone on the right if the histogram and the second is a 1ev increment display across the bottom. The 1ev increments let you see how many stops of dynamic range are contained in the image (ignoring the ERI zone).

First is ERI. Kodak describes ERI as a feature that allows overexposed details to be recovered using the RAW converter and negative exposure compensation. Basically what Kodak does is reserve some of their 12-bit resolution for ERI by defining less bits to be available for the output JPEG. The image review display clips if the exposure moves into the ERI range.

What I've read is that two bits of the twelve bit RAW range are reserved for ERI processing. That would imply that the true ISO of the camera is two stops slower than the rating. I set out to verify that using a strobe and Ike light meter. I set my camera and meter to ISO 80 and 1/125, adjusted my strobe power to read f/22 on the meter, then took an exposure at f/22. Rather than see a full normal exposure or a full ERI exposure, what I saw was an exposure that was half way into the ERI zone. It appears to me that the camera claims that the image is 1 stop overexposed but in reality it was 1 stop underexposed. Although my test was not rigorous, it suggests the true ISO rating of the 14n's sensor may be closer to 50 than 80. In order to shoot with the 14n taking full advantage of its sensor range you need to overexpose the image by 1 to 1.5 stops.

I then loaded the overexposed images in Photoshop CS to test the RAW converter and attempt to verify the ERI function. The ACR 2.0 histogram displayed no ERI zone but clipped the overexposed areas instead. By setting brightness to 0 and exposure to -1ev, the clipping was removed and the detail restored. ERI was definitely functional and the Kodak converter simply redefines histogram max and 0ev exposure. I also verified that the maximum ERI range was 1.5 to 2 stops. Great stuff as long as you understand it!

Equally interesting is the fact that the 14n displays 8 to 8.5 stops of dynamic range on its histogram display and the stops have irregular spacing (suggesting that the camera tries to apply DNR curves to the image to compress the image to fit in a JPG). The ERI zone apparently adds 1.5 to 2 stops of additional range resulting in a total DNR of 10 stops! It's not clear if the lowest stops are usable due to noise, but I've read that the camera is capable of consistently delivering 8 stops of range or more, making it a stop or more better than other digital SLR's available today.

So armed with this knowledge, I've been thinking about the 14n's much maligned noise performance. First, the 14n displays and unusually wide dynamic range within its histogram so its unfair to complain of excessive noise in a range other cameras simply clip off. Second, the Kodak ERI function (which amounts to a crutch for photojournalists) encourages shooters to underexpose their images by lying to them about the result. ISO 80 shoots more like ISO 160 but is really more like ISO 50. Clearly if you spot the 14n's competition 2 stops of proper exposure advantage, the 14n's noise may appear to suck. I now suspect that the 14n offers good noise performance as long as you expose properly and only use RAW.

The more I look, the more I suspect the 14n might make an excellent housed SLR. It offers incredible full-frame resolving power and up to 10 stops of dynamic range while using Nikon lenses. Now if only we had a housing...

Sweet! I would love to have 2 stops more of dynamic range. It will be VERY interesting to see how the camera handles "sunball" exposures.

Cheers

James

I don't think you'll get 2 stops more dynamic range vs other DSLR's. From Craig's discription, I think Kodak shot themselfs in the foot with their underexposure gimmic. They spotted their competition 2 stops without giving a good explanation and got written up for having much worst noise performance. They probably would have gotten better press if they had not done this and instead have competitive noise level, which I think they should be able to do with that sensor. That probably would have made this a more popular camera to date.

IMHO this is a really stupid idea considering their customers are likely to be very competent photographers who are generally capable of properly exposing their subject, and who's also heard of Photoshop and can intentionally underexpose and adjust if that was what he wanted.

Craig, how about this housing:

http://wetpixel.com/forums/index.php?showtopic=3715&hl=

I think Kodak shot themselfs in the foot with their underexposure gimmic.

Having done business and worked with Eastman Kodak over the past few years, this appears to be their business plan (shooting themselves in the foot that is)

IMHO this is a really stupid idea considering their customers are likely to be very competent photographers who are generally capable of properly exposing their subject, and who's also heard of Photoshop and can intentionally underexpose and adjust if that was what he wanted.

And again....

*sigh* Once agan Kodak manages to screw it up. No big suprise there.

M

I think the same thing. Their ERI function (not user defeatable) puts them at a 2 stop disadvantage right off the bat. I've been going back through the reviews looking for mentions of this but haven't found any. That leads me to believe that they were screwed by it since reviewers shot the camera as Kodak intended.

If you know about ERI and don't require JPG's looking good out of the camera, you can intentionally "overexpose" to get those stops back. Just set your RAW converter default to -2ev and ignore the ugly thumbnails. The downside is that the camera becomes exceptionally slow, perhaps even ISO 20! Z220's may be marginal.

I don't know about its ultimate dynamic range potential. Recent data I've seen suggests that the 14n has good noise in well lit scenes at lower ISO's. Thom Hogan claims that the 14n has more than a stop better dynamic range than anything else (and he hasn't added the 2 stops back, either). That would suggest that the 14n offers much better range than anything else, but I'm with Herb on that. It's hard to believe Kodak would have 2 or more stops better range than anybody else and maintain good noise immunity.

The 14n has show-stopping resolving power, better than literally anything else provided you avoid aliasing. If its noise performance is good and its dynamic range even a little better than others, it would have to make a great underwater camera.

Cruising the Rob Galbraith 14n review, i found this quote:

Kodak's highlight recovery magic is a result of the how the data coming off the sensor is processed at a given ISO, and how that effects the image's dynamic range. In short, the 14n depends on a combination of both analog gain (applied as the image data is converted to 0s and 1s) and digital gain (roughly akin to a Curves adjustment in Photoshop) to bring the image up to the level of the ISO set on the camera.

 

This is true regardless of ISO. For instance, with the camera set to ISO 80, the analog-to-digital conversion circuitry in the CMOS sensor pumps out image data with about an ISO 64 rating, says Kodak's Steve Noble. Digital gain applied to the image data during processing in Photo Desk, or in the camera out to ERI-JPEG, makes up the 1/3 stop between ISO 64 and ISO 80. Combine this with the dynamic range of the sensor and you're left with sufficient highlight headroom to rebound from fairly significant overexposure error.

 

Unlike previous Kodak digital SLRs, which depended on digital gain exclusively to achieve sensitivities higher than the base ISO, the 14n sensor utilizes analog gain at ISO 80, 160 and 400. Since analog gain and digital gain combine to give the camera's ISO range of 80-400 in 1/3 stop increments, the ISO settings at which digital gain is providing the biggest boost are the ones where overexposure recovery should be the most effective. In the case of the 14n, that's ISO 320 and ISO 125, or just shy of the upper two analog gain points.

First, it's my understanding that all cameras control their ISO rating through analog gain. Otherwise they're just throwing away bits in the ADC's. It is possible that a camera may have fewer analog gain increments than ISO settings. If that's so, a user would like to know so he can avoid the ISO settings that rely on evil digital gain.

Second, the suggestion that ISO 80 is really ISO 64 cannot be right. The 1/3 stop difference doesn't come close to the 2 stop capability that their ERI apparaently provides. 2 stops of recovery headroom means that ISO is really ISO 20, not 64. Perhaps these numbers are based on midrange sensitivity relative to film, but all that ultimately matters is knowing how to expose properly and not waste dynamic range.

Anyway, this seems to incrementally confirm what the 14n is doing. I haven't found anyone yet that suggests the 14n be operated in apparent overexposure to maximize its performance.

I'd like to know what the real noise performance of the camera is at its various "optimal" ISO settings. Perhaps a setting of 160 in combination with exposing at ISO 50 or 64 would yield best results without taking huge strobes or giving up optimal apertures for macro.