Amazon Kindle Unlimited Offer: 1-Month For FREE!

Need a new camera

Attention!

This topic is locked.

Reason : op questioned answered


9 Jan 2014 9:19PM

Quote:... Nowadays AF sensors are so good that they can work with low light and small apertures and modern EVFs can be set up to display whatever brightness you need. OVFs don't tend to work too well in the dark, of course, but I don't think a larger aperture will help much. ...


The latest generation of sensors is amazing, true. However, when taking night images, I found pretty fast that the sensor outstrips electronic viewfinder ( or camera screen in live view) abilities. A lot.

When going after a starry sky shot EVF it is hopeless - it's resolution does not allow to see smallest of the stars, only the major ones. That's not enough - as large pointy light sources are not great targets for manual focusing. Autofocus fails here as well. And low magnitude stars are smaller than the viewfinder's pixel - therefore unseen.

Another trap to fall in the night is long exposure work. The sensor still reads it OK, but EVF ( which is in fact video, not photo) shows an image either too dark, or too bright ( with brightness boost) to work comfortably. Definitely not WYSIWYG or not real time when shutter speed falls below 1/25c.

Should I mention that optical viewfinder is free of these problems? And yes, fast lens helps a lot - simply because an aperture two times larger ( say f/2 instead of f/4) gives an image four times brighter. That's physics law - and my personal experience to thatSmile

It is possible to theorise about anything, but only practice offers true answers. Therefore - try before buying into it. Camera makers cannot be too silly offering even their premium compacts with f1.8 lens.
Nigeve1 8 1.4k 101 United Kingdom
9 Jan 2014 10:46PM

Quote:You say that you want good low light performance, so you need the high ISO performance to be good. The difference here between these two is minimal. The 7100 has a much newer generation sensor and you would therefore expect it to be a lot better in this respect, the reason it is not is that 24MPixels have been crammed into the same sensor area compared to 16.1MPixels on the D7000. The individual pixels are smaller on the 7100 and that makes for worse relative high ISO performance

That's not entirely true nowadays. It used to be true because cramming in more pixels meant that there was more "dead space" on the sensor and also because each individual pixel generates a certain level of standing noise, so the more you have the more noise you get. Modern technology has reduced both these problems very significantly - to the extent that it doesn't make much difference (except at very high ISOs anyway).

However, the way that most review sites compare the noise from different cameras is to look at the image in native size. Because the 24Mp picture is larger than the 16Mp picture you see more noise - for obvious reasons. If you resize the pictures to the same size the noise will be very similar. I remember DPR pointed this out when they tested the 24Mp Sony sensor (A77) against the similar 16Mp Sony sensor (A55). The A77 seemed noisier until you resized the images, at which point they found the noise to be "effectively the same". I would have thought that the D7100 would be better than the D7000 under the same circumstances but I haven't seen any tests on this. The D7000 is a bit of a bargain nowadays though.



I am sure that technological strides have certainly been made in this direction, but I still think pixel density is a significant factor, I take on board what you say about the larger image size making noise more noticeable. This can be observed on the Imaging Resource website Comparometer by say comparing the D7100 with the D610. I suggest these as they have a similar generation of sensor and are both 24MP so the image size is identical at the native size. If ISO 3200 or 6400 are selected as quite but not excessively high ISO settings, there is a very clear advantage to the 610 over the 7100 in noise levels. So far as I can see here, the only significant difference between these two is the lower pixel density of the full frame 610 sensor compared to the APS-C 7100 sensor. This suggests that if Nikon had stayed around 16MP for the D7100 then with the newer sensor technology, it would have significantly better high ISO performance than it does with the 24MP sensor.
DXOmark have compared the D7000 and D7100 and quote a similar image quality in terms of noise for the D7000 at ISO1167 compared to the D7100 at ISO1256 which is all of 0.1 of a stop difference and from a practical point of view negligable.
Steppenwolf 9 1.2k
10 Jan 2014 8:41AM

Quote: This can be observed on the Imaging Resource website Comparometer by say comparing the D7100 with the D610. I suggest these as they have a similar generation of sensor and are both 24MP so the image size is identical at the native size. If ISO 3200 or 6400 are selected as quite but not excessively high ISO settings, there is a very clear advantage to the 610 over the 7100 in noise levels.


My point was that the pixel size doesn't make much difference to noise in a modern sensor - within limits of course. The sensor size obviously makes a big difference. An FF camera has double the light collecting area so it will give a 1 stop advantage over APS-C, everything else being equal.
Nigeve1 8 1.4k 101 United Kingdom
10 Jan 2014 1:42PM

Quote:
My point was that the pixel size doesn't make much difference to noise in a modern sensor - within limits of course. The sensor size obviously makes a big difference. An FF camera has double the light collecting area so it will give a 1 stop advantage over APS-C, everything else being equal.



Thanks for that clarification, Steppenwolf. I would be interested to know on that basis what you think the result would theoretically likely to be if you were to compare a D610 used in DX mode against a D7100, obviously with lens and all other factors the same. Would there be likely to be a significant difference in high ISO performance given that the used sensor area is then effectively the same.
Steppenwolf 9 1.2k
10 Jan 2014 2:05PM

Quote:
Thanks for that clarification, Steppenwolf. I would be interested to know on that basis what you think the result would theoretically likely to be if you were to compare a D610 used in DX mode against a D7100, obviously with lens and all other factors the same. Would there be likely to be a significant difference in high ISO performance given that the used sensor area is then effectively the same.



If you used a D610 in DX mode you'd end up with a 12Mp picture, of course. If you examined that picture against a D7100 picture (which had been resized down to the same size) I'd expect the noise levels to be pretty similar except at really high ISOs (e.g. ISO25600). The lower density pixel D610 would still be slightly better at more normal ISOs but not so much that the user would notice without special equipment. When you get up to the very high ISOs in low light the standing noise of each individual pixel becomes more and more significant.

If you didn't resize the D7100 image, of course, then it would look much more noisy than the D610 DX image - about a stop worse. But the picture would have double the area, so that's not unexpected.

You'll find loads of stuff on the web about this e.g.. There have also been loads of debates on here (mostly locked of course - what a surprise!) and DPR over the years.
ChrisV 14 2.3k 26 United Kingdom
10 Jan 2014 2:46PM
The important caveat to this is of course 'the very high ISOs' because the more you reduce the size of the individual sensel and boost its sensitivity [ie you effectively half the sampling rate of the photons hitting the site - this is shot noise as opposed to the denigration of the signal due to imperfections in processing, heat transfer and so on [which has been continuously improved since the advent of digital sensors].

I don't claim to fully understand the physics or know what the absolute limits are, but it underpins my understanding of why Nikon and Canon have limited their high-end sensors to 16 and 18mp respectively - in order to offer sensitivities as high as 204,800 [although those settings look unusable to me and I can't see the D4 has made any significant improvement over the D3s at it's maximum of one stop less]. The problem is of course that once you lower photon collection beyond a certain level [which is area x amount of photons] you come to a limit whereby because photons are completely random, the data collected is unreliable. Although you can eliminate noise by averaging neighbouring sensels [the trick that 42mp Nokia employs - 'pixel binning'] you are rendering their results as individual points null [which is why resolution tends to dip as you boost sensitivity in all cameras].

Look at it like measuring rainfall - if you want to know its rate over a large area, having a lot of little buckets will give you a reasonable indication. If you want to measure very specifically, the larger the bucket you have, the more reliable will be your measurement. If you have a tiny bucket and very light rainfall, you may collect a drop or alternatively not get anything, [which is analagous to a small photosite and very low light]. In such situations what you collect is useless because it gives an all or nothing reading. That is essentially the reason photosites are not further multiplied and why sensitivity boosting can only go so far.
Steppenwolf 9 1.2k
11 Jan 2014 8:31AM

Quote:The important caveat to this is of course 'the very high ISOs' because the more you reduce the size of the individual sensel and boost its sensitivity [ie you effectively half the sampling rate of the photons hitting the site - this is shot noise as opposed to the denigration of the signal due to imperfections in processing, heat transfer and so on [which has been continuously improved since the advent of digital sensors].


Yes, but shot noise can be effectively eliminated by pixel binning as you suggest - theoretically at least. For example, if you take the case of a camera which is available in 12Mp and 24Mp versions with the same sensor size, the 24Mp version will suffer more from shot noise because its sensels are half the area and therefore subject to more random variation in photons collected than the larger sensels. So, if you choose to display your 24Mp image at 100% it will appear more noisy than the 12Mp image - but you've got more resolution. If you display your 24Mp image at the same size as the 12Mp image (i.e. at the same resolution) the extra pixels of the 24Mp sensor are binned together and the noise will reduce to the same level as that 12Mp sensor - but you will have lost your resolution advantage of course.

The extra pixels are basically a "free lunch". If you need the extra resolution you have access to it but the price you pay is higher noise (which may be insignificant depending on the lighting conditions). If you don't need the resolution then the picture can be down-res'ed and you've got the 12Mp resolution and (to a large extent) the 12Mp noise. The reason that low Mp cameras are still better at very high ISOs is not related to shot noise - it's related to the electronic imperfections of the sensors not quantum limitations. If the higher Mp camera had a "perfect" sensor then pixel binning would perfectly replicate the output of the lower Mp camera.
strawman 17 22.2k 16 United Kingdom
11 Jan 2014 10:56PM

Quote:The extra pixels are basically a "free lunch"
As we have discussed many times "not true for all aspects of noise." a lot of silicon noise come from the boundary interfaces, so add up all the edge areas of the pixels and the related circuitry and you find the higher resolution sensor has more boundary area so those noise factors become more apparent. How big an impact depends on the technology used. Honest it is there.
Steppenwolf 9 1.2k
12 Jan 2014 8:20AM

Quote:The extra pixels are basically a "free lunch" As we have discussed many times "not true for all aspects of noise." a lot of silicon noise come from the boundary interfaces, so add up all the edge areas of the pixels and the related circuitry and you find the higher resolution sensor has more boundary area so those noise factors become more apparent. How big an impact depends on the technology used. Honest it is there.


I'm well aware of that SM - which is why I said in my first post "It used to be true (that noise increased with number of pixels) because cramming in more pixels meant that there was more "dead space" on the sensor and also because each individual pixel generates a certain level of standing noise, so the more you have the more noise you get". I know you like to nit-pick, but please read the post.

This "standing noise" - if I'm allowed to call it that - is the reason why Canikon make low Mp cameras for those who want the best performance at the highest ISO levels. It's not because of "shot noise". If all the sensels on a sensor were "perfect" and simply accurately counted the number of photons landing on them there would be no need to make these cameras.
NikonNovice 14 107 1 United Kingdom
12 Jan 2014 9:25AM
I bought a D7100 and shortly after a Tamron 24-70 2.8 lens because it has anti shake built it whereas the Nikon version does not. I would say dont get too hung up on noise as I have seen older camera shots which look stunning and the noise was taken out in software. Buy the D7000 or if you can afford it the D7100 and by the time you notice any (and I do mean any) noise disadvantage you will be ready to move on to a bigger, better, newer and more expensive camera anyway. Any of these two cameras will put you on the learning curve to be producing stunning shots long before you notice any limitations. You really cant go wrong with any of these two. Improve your technique over the next few years and see what new shiny cameras are out there to buy then.
lemmy 14 2.9k United Kingdom
16 Jan 2014 11:31AM
I think I may have found the answer to the 'need/ want a new camera' syndrome that affects us all.

I've just adapted my laser printer for 3D printing. Here's my first shot at using it. The camera looks great and works just the same as the real thing but I'm feeling a bit fed up because it simply will not print out the lens cap no matter what I try.

It's typical, you think you have saved yourself a load of money and then you end up having to fork out 7.50 to the camera maker anyway Sad

3dprint.jpg
ChrisV 14 2.3k 26 United Kingdom
16 Jan 2014 1:35PM

Quote:The extra pixels are basically a "free lunch" As we have discussed many times "not true for all aspects of noise." a lot of silicon noise come from the boundary interfaces, so add up all the edge areas of the pixels and the related circuitry and you find the higher resolution sensor has more boundary area so those noise factors become more apparent. How big an impact depends on the technology used. Honest it is there.

I'm well aware of that SM - which is why I said in my first post "It used to be true (that noise increased with number of pixels) because cramming in more pixels meant that there was more "dead space" on the sensor and also because each individual pixel generates a certain level of standing noise, so the more you have the more noise you get". I know you like to nit-pick, but please read the post.

This "standing noise" - if I'm allowed to call it that - is the reason why Canikon make low Mp cameras for those who want the best performance at the highest ISO levels. It's not because of "shot noise". If all the sensels on a sensor were "perfect" and simply accurately counted the number of photons landing on them there would be no need to make these cameras.



Your standing noise sounds a lot like my shot noise - except for the fact that 'standing noise' is your own invention. The fact is the light sensitive area of any sensor must decrease as the sensel count goes up. That's because sensels have boundaries and those boundaries account for a proportion of the overall area occupied. Now as for averaging out noise by pixel binning, I have no argument that that can be very effective up to a certain level. Once you go beyond the tolerance where your reading becomes so unreliable that it isn't even very useful taken as an average amongst its neighbours, you are in trouble - or "screwed" - if I'm allowed to call it that.

Alternatively it could be that the engineers at CaNikon/Sony just don't like giving out free lunches. Even when you've paid through the nose for them.
16 Jan 2014 7:58PM
To add to the points above, increase in pixel numbers changes not only noise level, but noise pattern as well - thus making it uglier and more noticeable as resolution grows. Where is acceptable balance here? There is no certain rule, as noise acceptability depends on image final presentation - which varies a lot. Technically pixel numbers must stop where sensor resolution becomes limited by optical resolution. This threshold is already reached for compact sensor cameras( that's why binning becomes so popular), there is little wiggle space for APS-C, and a lot of potential in full format.
Steppenwolf 9 1.2k
17 Jan 2014 8:34AM

Quote:Your standing noise sounds a lot like my shot noise - except for the fact that 'standing noise' is your own invention.


No, it's fundamentally different. What I mean by the term "standing noise" is the noise generated by the imperfections in any electrical device - what SM calls "silicon noise from the boundary interfaces". You find this in any analogue device. For example the more devices you add on to process the signal from an electric guitar the more the noise that gets generated - each device adds its own characteristic signature. In the case of sensors the technology has been improving to reduce this level significantly over the years.

Shot noise is the result of the quantum nature of light and can't be reduced. The more resolution you extract from a sensor the more shot noise you'll see - it's one of those things that are inherent in dealing with photons, like diffraction.


Quote: Technically pixel numbers must stop where sensor resolution becomes limited by optical resolution.


That's not true according to Prof Bob Newman (who writes in AP). It would take too long to explain - and would probably cause a lot of acrimony.
ChrisV 14 2.3k 26 United Kingdom
17 Jan 2014 10:20AM
If you meant thermal, or Nyquist noise why didn't you say so? These things can and have been tackled and improved to massive degrees over succeeding sensor generations - it's arguably the main reason along with improving algorithms that we've seen such dramatic strides in noise control.

However, from purely empirical observation, I don't think there has been any significant step forward in the noise levels at the highest ISOs in the four years since the appearance of the D3s. [That is that at a similar size of reproduction I do not see significantly less noise in images produced by the D4/D1x at ISO 102400 than the D3s - their one stop higher ISO boost levels are pure marketing].

So if we haven't hit a limit in terms of what can be done, we do at least appear to have hit a road block. Now whether that is shot noise caused by the random nature of photon flow, or noise caused by interference from circuitry I suspect neither of us is qualified to say. One thing which appears logical is that after a certain level of signal boost it appears counterproductive to have a large number of photosites. My gut feeling is that is that when dealing with extremely sparse and sporadic photon collection it is better to have larger individual sites. You on the other hand seem to be saying that it is the increased circuitry of a larger number of sites which causes significant enough interference to cloud the lower level of signal present.

As I say - I concede I don't know enough to say which of these scenarios [or whether or not it's a combination of the two] is actually the cause. Nevertheless it appears fairly indisputable that at a certain low signal threshold, reducing the number of photosites does improve the chances of capturing a useful image.

In good light there may be no such problem. But if we could guarantee good light we would not need large photosites, not need large sensors and do away with large lenses with their wide apertures. [For DoF control there's always the Lytro light ray technology for post capture focusing].

Therefore whichever way you slice it, at the upper end of sensitivity there very clearly is 'no free lunch'.


Quote:Shot noise is the result of the quantum nature of light and can't be reduced. The more resolution you extract from a sensor the more shot noise you'll see - it's one of those things that are inherent in dealing with photons, like diffraction.


Good to see you are now accepting that shot noise cannot be got rid of anyway - strange however that you don't seem to think it has any implication for the scale of photon collectors - or do you?