If Nikon doesn't get their act together on DX...

Started Feb 7, 2013 | Discussions thread
jfriend00
Forum ProPosts: 11,131
Like?
Re: Who are "we"...
In reply to Kerry Pierce, Feb 9, 2013

Kerry Pierce wrote:

jfriend00 wrote:

altair8800 wrote:

If you want the same noise on DX as FX at the same shutter speed, you need a wider aperture on DX.

You have to take pixel density into account. In the situation where the FF has fatter pixels, this is true. If the FF has 28Mp for same pixel density as D300 and same level of efficiency, then same noise with same lens.

OK, we're mostly in agreement, but this part of your supposition isn't right. Noise in an overall image is related to total photons collected not to pixel density. And an FX sensor collects way more photons than a DX sensor (because it's larger) and thus delivers lower noise.

I hate to be disagreeable, but I can't agree with this. This is part of the "FX is always best" propaganda. The size of the sensor doesn't matter for noise performance. Pixel density and the state of the technology, are what matters.

If it were only sensor size that determined noise performance then, of course, medium format sensors would be the kings of high ISO cameras. FF for Nikon users goes back to the Kodak SLR/N days. Nobody ever said the SLR/N was a high ISO monster. Seriously, all you need to do is compare the performance of the various FF sensors over the years. Forget the DX vs FX thing and simply look at the older FF cameras vs the new generations. The size of the sensor is identical, thus not a factor, yet the noise performance is different with each generation, because of technology.

OTOH, if you want identical performance for DX and FX, all you need to do is cut a DX sized sensor from the same wafer used to make the FX sensor. You could certainly have a DX sensor with the same noise performance of the d3s or d4, but you would have to live with the much smaller amount of MPs for the DX sensors, probably about 6mp or so, for a DX version of the d4.

You may be talking about per-pixel noise (where fatter pixels would appear to be better),

Per pixel noise performance is an important metric, because that is the limiting factor for determining the amount of cropping one can do and the limiting factor for the size of a given print. For someone that never crops and never prints larger than 8x10, maybe per pixel noise isn't important. I don't know because I don't know anyone that doesn't do those things.

But, anyone that preaches the use of the d800 instead of a DX camera, has to be concerned with per pixel performance because they are cropping an incredible amount, from 36mp to 15mp....

but per-pixel noise is not an important characteristic in the visual perception of noise in a viewed image. If that was the case, then the D800 would have much worse noise than the D4 (because it has much smaller pixels than the D4 and worse per-pixel noise than the D4) but, the D800 does not produce images with worse visual noise in them. Further, fatter pixels make more visible noise. In fact, the impact of the D800 per pixel noise is reduced because it has such tiny pixels so the noise artifacts are smaller and more random and thus less visible in a printed or screen viewing of an image.

As an owner of both a d3s and a d800, I would again have to disagree. The d3s has wonderfully fat pixels and it certainly doesn't suffer because of that fact. The d3s can produce usable images from insanely high ISO levels, long after the d800 files have turned to mush. IME, so long as you are severely restricting both the viewing/print size and the cropping depth, then the d800 apparent noise at the higher ISO's isn't too bad. In that respect, it's true that the higher number of pixels are helpful, especially if you're willing to do more in PP. I haven't done a side by side test of the d700 and d800, but my impression is that the d800 does well in that comparison. That, in itself, is a remarkable achievement, IMO.

I can't give precise instructions and examples because I simply don't use the d800 at high ISO and don't have to worry about fixing any noise problems.

But, this will be an important metric if Nikon gives us a 24mp d400. Having briefly looked at the DPR samples for the new Toshiba sensor in the d5200, my impression is that it has similar noise performance to the d7000. Take a look at DPR's Studio Shot comparison where I have the RAW d5200 compared to the d7000, d800 at ISO 200 and then the d3s at ISO 1600.

I'd need to download and play with the d5200 NEF's before I'd make a final judgement, but my initial impression is that the sensor might be okay for a d400, especially if you're willing to downsize significantly. For typical night sports shooting, that might be an acceptable solution. I don't yet know how that would shake out. If I get time, maybe I'll do a test series with the d800 and see how much extra work is involved in processing low light shots at high ISO.

Kerry, I'm not dissing the D3s or D4 performance - they are phenomenal. I was trying to correct a wrong impression by one particular poster that the size of the pixels is THE determinant of noise and fatter pixels are always better. If that was the case, we'd have 4MP cameras that were running circles around the D3s and that isn't the case. It is not only the pixel size that matters. There are a whole bunch of interrelated characteristics. Adding more pixels makes more per-pixel noise (lower S/N in each pixel) and complicates edges losses and complicates read-out, but more pixels also makes individual noise artifacts smaller and more averaged/self cancelling. There are lots of individual tradeoffs here and it is NOT the case that fat pixels are always better than smaller pixels. Visible noise in a printed or screen viewed image is a system issue, not just a factor of how big the pixels are. That's what I was trying to explain.

-- hide signature --
Reply   Reply with quote   Complain
Post (hide subjects)Posted by
(unknown member)
noNew
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark post MMy threads
Color scheme? Blue / Yellow