 |
 | Home | Contact Us | FAQ | Search & Site Map | Link to Us |
 | Sign In | Join | Other 45 Sites in Network |
| PhotoKB HomeDiscussion Groups Digital Photography Digital PhotoDSLR CamerasZLR CamerasPoint & Shoot CamerasFilm Photography 35 mmLarge FormatMedium formatDarkroomFilm and LabsOther EquipmentPhoto Technique Nature PhotographyPeople PhotographyTechnique GeneralGeneral Photo Topics General TopicsAustralian PhotographyUK PhotographyDirectoryPhoto Clubs |
Photo Forum / Digital Photography / DSLR Cameras / November 2006ALMOST!!
http://www.flickr.com/photos/pete_with_pentax/285444496/ Some have got them in the US, but not me yet, maybe next week. Afterall, Australia is a long way from the Pentax factory. :-( > http://www.flickr.com/photos/pete_with_pentax/285444496/ > > Some have got them in the US, but not me yet, maybe next week. > > Afterall, Australia is a long way from the Pentax factory. :-( A big disappointment from the samples I've seen so far. Noise is way too high even at iso 400. I've yet to see one single image that could genuinely be described as better than one from an ist D or DS. Remember the Slade hit from the seventies? 'C'mon feel the noise'?, maybe it should be the new Pentax theme song. > A big disappointment from the samples I've seen so far. Unless you are talking about images posted in the last 48 hours or so since the camera started shipping, you are looking at images shot with pre-release firmware. All reports on dpreview - including those accompanied by images shot with the production model camera over the last couple of days - show a quite noticeable improvement. One post on dpreview claims to have done objective statistical measurements of the noise levels (computing standard deviations of luminance levels between pixels on flat dark backgrounds, etc), and reports that the images shot with the actual released firmware are something like twice as as good in this department as the pre-release ones. Definitely better than the Sony Alpha, and arguably better than the D80, although I gather the D80 did some in camera NR even on its RAW images to get that result? > I've yet to see one single image that could genuinely be described as > better than one from an ist D or DS. "Better" in terms of noise, no, that would be unreasonable to expect. For the same sensor size, a 10MP sensor cannot perform as well as as 6MP one in this department - the pixel sites are not as large, and hence have less data to play with for a given exposure. Apparently we are talking on the order of 1/2 stop of difference - noise at 1600 on the K10D is worse than that of the DS at 1600 but better than that of the DS at 800. Whether or not this increase is noise is offset by the increase in resolution (or the myriad of other improvements in the K10D) is, of course, a personal decision. Frankly, for me, it is not. I've yet to feel limited by resolution on my DS, but often wish I could get less noise at ISO 1600. So between that and the size/weight issue, I'm hoping to eventually see something like the K100D but with just enough of the upgrades from the K10D so tht it doesn't feel like I'm stepping back from the DS in most ways just to get SR. > Remember the Slade hit from the seventies? 'C'mon feel the noise'?, > maybe it should be the new Pentax theme song. Not just Pentax - again, higher noise at 10MP than 6MP for the same sensor size is a simple matter of physics. But if the numbers above are to be believed, Pentax is doing better than the others in this department. --------------- Marc Sabatella marc@outsideshore.com Music, art, & educational materials Featuring "A Jazz Improvisation Primer" http://www.outsideshore.com/ >> A big disappointment from the samples I've seen so far. > [quoted text clipped - 3 lines] > accompanied by images shot with the production model camera over the last > couple of days - show a quite noticeable improvement. It is the DPR posts that I based my judgement on. It's all subjective, but for me noticeable noise as low as iso400 is not what I want. I've now cancelled my K10 order and will probably sell my pentax gear and move to Canon - no point in buying a D200 or D80 as it's bedevilled by the same sensor. That's just my own thoughts, others are, of course, free to be happy with the K10 if that's how they feel. A K10 + grip + battery + additional flash to compensate for the lack of body wireless sync would see me shelling out over another 1K, and I had to ask myself whether, after ploughing yet more money into Pentax, I'd be happy with the results from the new camera - and, in my heart, I know I won't be. So, time to bite the bullet and sell up and move on. (I've got a couple of nice Limiteds, if you're in the market) > Not just Pentax - again, higher noise at 10MP than 6MP for the same > sensor size is a simple matter of physics. But if the numbers above are > to be believed, Pentax is doing better than the others in this > department. And yet, the noise that I, at least, object to is at low illumination levels and in the shadows (I don't really care about noise in areas with high illumination at eg ISO 1600, since it is actually very little in current cameras). This has little to do with the pitch size or full well capacity, and more to do with the magnitude of read noise. This is why Canon cameras do so well at high ISO/low light photography; see Roger Clark's measurements of this at http://www.clarkvision.com/imagedetail/digital.sensor.performance.summary/index.html You can also see that this read noise decreases quite a bit as ISO is incresed for the Canons, not so much for cameras with Sony CCDs. This kind of noise isn't really correlated to the pixel size, so could in principle be improved in future cameras with the same pixel pitch as the current 10mp sensors. > And yet, the noise that I, at least, object to is at low illumination > levels and in the shadows (I don't really care about noise in areas > with high illumination at eg ISO 1600 I don't know what kind of "low light" shooting you do, but I am often shooting in settings where even with <2.8 apertures and shutter speeds at the limits of what is needed to eliminate subject motion blur, ISO 1600 still produces images with very low illumination levels. > This has little to do with the pitch size or full well capacity, and > more to do with the magnitude of read noise. This is why Canon cameras [quoted text clipped - 3 lines] > You can also see that this read noise decreases quite a bit as ISO is > incresed for the Canons, not so much for cameras with Sony CCDs. I can see how this would be true, and I should have qualified my statement about "small pixels = more noise" to refer only to sensors using essentially the same type of technology (CCD/CMOS, etc). I don't think anyone was expecting Pentax or Nikon to suddenly make such a fundamental change to their cameras at this point, so again, expecting either to come out with a 10MP/APS-C camera with less noise than their own 6MP/APS-C cameras would not have been reasonable. --------------- Marc Sabatella marc@outsideshore.com Music, art, & educational materials Featuring "A Jazz Improvisation Primer" http://www.outsideshore.com/ >>And yet, the noise that I, at least, object to is at low illumination >>levels and in the shadows (I don't really care about noise in areas [quoted text clipped - 4 lines] > at the limits of what is needed to eliminate subject motion blur, ISO > 1600 still produces images with very low illumination levels. Well, my "low light" shooting tends to be exposures in the minutes, but that is besides the point :) OK, it seems you are saying that you end up having to essentially shoot at ISO 3200 or 6400 or whatever. In that case, it is even more important to have good shadows (since you're basically using only the "shadow" region). The actual light level is irrelevant, the only thing that matters is how many photons were collected in total, and the total magnitude of the read noise. Of course, a larger pixel will end up collecting more photons per unit time for the same scene and f/stop, as you said. But look at the Canon 20D and the Nikon D200: The 20D really is better at high ISO (I used a 20D for some time to see how I liked it, but got a D200 in the end; there is no question about it, though, the 20D has better shadows at high ISO). Anyway, the pixel size is almost the same (or look at the D70, it is almost the same with the D200, and has bigger pixels than the 20D). And if you actually look at the differences, they are in the shadows. There is no difference in midtones and above. The reason is the difference in the read noise, I think. Look at the read noise Clark measured for ISO 1600: In terms of recorded numbers, read noise is 15 DN for the D200 but only 5 for the 20D. Of course, this is partly because the 20D is more efficient at detecting photons (slightly higher unity gain), but the main reason is lower read noise. >>This has little to do with the pitch size or full well capacity, and >>more to do with the magnitude of read noise. This is why Canon cameras [quoted text clipped - 11 lines] > either to come out with a 10MP/APS-C camera with less noise than their > own 6MP/APS-C cameras would not have been reasonable. But look at Clark's measurements for the S70: Read noise is also at around 3 electrons at its unity gain (which is at ISO 100). It uses a CCD. So whatever method Canon is using to reduce read noise is certainly possible with CCDs, and not expensive. It does not seem to be tied to the CMOS sensors, although I may be missing something. Anyway, to summarise: My point is that significant reduction in apparent noise is possible by lowering read noise, which Canon seems to achieve even in its compacts. > Anyway, to summarise: My point is that significant reduction in apparent > noise is possible by lowering read noise, which Canon seems to achieve > even in its compacts. Perhaps I should have mentioned that I find the low read noise of the S70 very strange (and since it presumably uses a Sony CCD, it must be due to what Canon's processing does to the signal afterwards, ie they have magical electronics), but assume Clark must have checked carefully once he found that result, so I have no clue what is going on there. > OK, it seems you are saying that you end up having to essentially > shoot at ISO 3200 or 6400 or whatever. Actually, no - my camera goes up to 3200, but I consider the noise and loss of detail unacceptable. So I shoot 1600 and live with shots that are a little n the underexposed side and still throw away a good many shots because of subject motion blur. > In that case, it is even more important to have good shadows (since > you're basically using only the "shadow" region). Right, and this is the situation I had in mind when I observed that, with all else equal, more pixels = smaller pixels = more noise. But it is true I was not considering the possibility of improving read noise performance, and I admit to not fully understanding the issues involved here, so I think I should just sum up by saying, while it might be *possible* for a single camera line to show lower noise despite a higher pixel count, it certainly shouldn't be *expected*, as unless there are other technological improvements made (that could presumably also be incorporated into the lower pixel count sensors), the default would be to get more noise. --------------- Marc Sabatella marc@outsideshore.com Music, art, & educational materials Featuring "A Jazz Improvisation Primer" http://www.outsideshore.com/ > > OK, it seems you are saying that you end up having to essentially > > shoot at ISO 3200 or 6400 or whatever. [quoted text clipped - 3 lines] > are a little n the underexposed side and still throw away a good many > shots because of subject motion blur. Yes, I realise that. I mean that this way you get results exactly the same as if you had used 3200 (say). If you use 1600 and underexpose by 1 stop, and then add 1 in raw conversion, I at least find that I get as bad (if not slightly worse) results as if I had used 3200 in the first place. If you do not add 1 stop, but leave it underexposed, noise is less visible (being in the shadows, which are now very dark). What I do the few times I need to shoot handheld and don't have enough light is to increase the blackpoint in Levels and also use curves to depress the shadows while enhancing the highlights. This doesn't work very well all the time (it depends on how underexposed it is, how much detail should be in the shadows etc), but sometimes it gives very usable results at high speeds. It's basically selectively adding exposure to areas where noise isn't visible. >> Actually, no - my camera goes up to 3200, but I consider the noise >> and [quoted text clipped - 10 lines] > place. If you do not add 1 stop, but leave it underexposed, noise is > less visible (being in the shadows, which are now very dark). Right, and this is pretty much what I do. I figure if the picture was *taken* in low light, it might as well *look* like it was shot in low light. So while I may apply a curve to brighten the lights some - much as you describe as well - I leave the darks more or less as they are. And I definitely find the results I get this way preferable to the results I get shooting at 3200 on my DS. But note I'm also one of those heathens who rarely shoots RAW. Perhaps I'd get better results at 3200 in RAW. But at 1600, I'm generally happy enough with the JPEG to not consider RAW worth the trouble unless I really have especially high hopes for the shoot. --------------- Marc Sabatella marc@outsideshore.com Music, art, & educational materials Featuring "A Jazz Improvisation Primer" http://www.outsideshore.com/ > with all else equal, more pixels = smaller pixels = more noise. But it > is true I was not considering the possibility of improving read noise [quoted text clipped - 5 lines] > incorporated into the lower pixel count sensors), the default would be > to get more noise. To some extent though... noise is at least less noticeable with more pixels. To the extent that noise reduces detail, which has to be pretty bad, more pixels add detail. |
Reply to this Message |
 Sign In
 Join
 My Latest Posts My Monitored Threads My Blog My Photo Gallery My Profile My Homepage Start New Thread Enable EMail Alerts Rate this Thread
|
©2009 Advenet LLC Privacy Policy - Terms of Use
This website includes both content owned or controlled by Advenet as well as content owned or controlled by third parties.