crashpc
Veteran Member
- With older Canon cameras, shooting with "BTTR" technique including max usable ISO setting, resulted with least noise. 6DII should still be plagued by this. Others are not that much.
-
Welcome to the new forums! Please read this first. For known issues we are working to resolve, click here.
Difference between ISO 100 pushed 6 stops and ISO 6400?
- Thread starter richarddd
- Start date
fishy wishy
Veteran Member
- Messages
- 9,358
- Solutions
- 4
- Reaction score
- 6,647
That's ironic, because I didn't know Sigmas had +-5 stops worth of iso rationally usable in them. 100, 200, 400 (and now where I saw Sigma IQ fall to bits) 800, 1600, 3200?
Mind you, I've seen people trying to 'prove' the opposite with banded iso 3200 shots in B&W.
John Sheehy
Forum Pro
EC adjustment is not just for indirectly getting a different ISO exposure index. It is also for getting the stated one for intended "gray" in scenes where the key and meter weighting would otherwise result in a different ISO exposure index.
The Sigmas get those higher ISOs by EC under the hood, anyway. Sigmas (at least the older ones) only have a single gain, both analog and digital, for all ISOs. IOW, if you take the same shot on a Sigma in M mode without changing Av or Tv values, starting at the highest ISO, down to the lowest, the RAW histograms are the same, and the read noise relative to saturation is the same (same DR) at all ISOs. Truly "ISO-invariant" in every sense.
What kind of bands?
Countour bands or line noise?
Contour bands come from over-quantization, which means that too much NR was applied for the file or display precision, in most cases, but it could be from actual quantization.
Line noise may present itself in the RAW in a way that helps one remove it. While RAW color is beyond the ken of most photographers, B&W is really an area where it is fairly simple for one to do their own digital darkroom work, if the three original RAW channels are just turned into a linear TIFF. You don't have to understand anything about the complexities of color (especially the Sigmas, which record something very far from RGB color) in RAW conversion to manually process RAWs for B&W.
fishy wishy
Veteran Member
- Messages
- 9,358
- Solutions
- 4
- Reaction score
- 6,647
Eh? Can you explain?
Last edited:
John Sheehy
Forum Pro
You said more, though, which isn't true. In any camera that has additive or offset read noise (what makes the recording of a lens cap wrong), the weaker the signal, the more it is obscured by read noise than decimated by photon noise. With every halving of signal, read noise doubles and shot noise only increases by 41%, relative to signal.
Yes, the photon noise of an extremely low exposure can have "problems" of its own (mandatory reality), but read noise (camera-added junk) can be a much bigger problem.
True, but only always true for the post-gain part of the read noise. The pre-gain read noise is fixed in a camera with no variable conversion gain. There are sensors now with changes in conversion gain up the ISO setting scale, and these cameras can drop absolute pre-gain read noise a full stop or more, just going from ISO 1600 to 2000, or 400 to 500. Such a camera can not be ISO invariant, and such cameras add a new level of complexity to the selection of ISO settings, both in manual ISO mode and auto-ISO with EC. Perhaps future cameras may have independent conversion gain for every ISO setting, and the whole idea of ISO invariance goes out the window, and we can only speak of it over specific ranges of ISO settings. If the changes were gradual, though, like every stop or even every 1/3 stop, it will make decision-making a bit easier, with smoother ramifications. Suddenly dropping of 1.3 stops of read noise at only one ISO sequence step might bring less absolute noise above a certain ISO setting, a pure empirical "gift", but it is a psychological monkey-wrench to someone who understands that and makes decisions.
That is not the precise statement that you might think. One help would be to keep post-gain read noise and pre-gain read noise distinct. When you are talking about things where they behave quite differently in your hypothetical comparisons, there is no such thing as "read noise".
That's a hint that you might mean post-gain read noise, but you didn't state that. At high ISOs, post-gain read noise (what you seem to be calling "read noise") is the smallest contributor to total read noise. You can't call that "read noise", as you did. That only creates confusion.
Last edited:
John Sheehy
Forum Pro
Sigma RAWs have to jump through hoops to get color out of them. They work on the principal that the red end of the spectrum penetrates silicon deeper than the blue end, so they do not have R, G, and B, but rather, R, R+G, and R+G+B, to put it in an oversimplification that makes the point. Spatial color resolution is where the Sigmas excel; they are actually poor at separating color within the color spectrum. If you look at a Sigma RAW directly where the top layer is represented as blue on your monitor, the middle green, and the bottom red, the images are very unsaturated, and can sometimes just appear to be almost a tinted monochrome if chromatic noise is not obvious.
fishy wishy
Veteran Member
- Messages
- 9,358
- Solutions
- 4
- Reaction score
- 6,647
Thankyou. Sigma Foveon represents a fringe club, and the only way to justify it as far as I can see is with high pixel counts that can at least offer something with truly high resolution even if it's only worthwhile at base iso, as with the SD1 15MP(x3). The confused Quattro devalues the concept because it takes a step backwards towards Bayer for the sake of marketing it seems. At least, way back with the SD10 I felt it was only realistically an iso 100-200 camera.
But come back to me if you see it any differently!
Last edited:
John Sheehy
Forum Pro
There are some freaky cameras out there that break your rule. I have one old superzoom Panasonic where the ISO range is 100, 200 ... 1600, and absolute total read noise is higher at 1600 than 100, highest at 800, and lowest at 200. It is amazing how pathetic a manufacturer can be, poisoning RAWs with unnecessarily noise so that intended white is the same RAW level in an ISO 1600 file as it is in an ISO 100 one, giving larger compressed RAWs to boot, with more bits of non-compressible noise, which come at the unnecessary cost of reduced headroom! I have an old Casio superzoom that goes to higher ISOs, but RAW is only 100 or 200. They might recognize the same problem with their electronics that was in the Panasonic, and limited the RAW ISO to 200, but they didn't give the user a way to get a usable preview image or default JPEG, shooting 3200 from an ISO 200 RAW; they are under-exposed by 4 stops. Manufacturers can be so obnoxious.
John Sheehy
Forum Pro
I found the Quattro concept to be a disappointment, because it was a step back in resolution of the middle and bottom layers. However, the concept has an advantage over the Bayer CFA. The Bayer CFA is completely red- and blue-blind to much of the surface area of the sensor, even when supplied with an AA filter, if the optics and stability are sharp enough. The worst that a Quattro can do is just have a very thin blind alley between pixels, as the next pixel samples the same color band, and that is pretty much filled in by diffraction, and by sensor diffusion at the lowest layer. So, the worst you get with the Quattro is a box filter.
So, even though the luminance tends to be more aliased in Sigma images, than AA-layered Bayer sensors, the red and blue channels as independent images, are more aliased.
For color resembling RGB, perhaps. The noise would be more tolerable if you added the three layers to make a monochrome. In fact, at least in the earlier models (SD10, I am sure), there was some zero-sum-game noise between the top and middle layers that disappeared completely if you blended them in the correct ratio. That's why color distinction in shadows between blue-greens was very bad, and gave large blotches in shadows in that color range, especially.
If you want to continue this Foveon color/resolution discussion, please start a new thread or find one already discussing it. Let's save these 150 posts for mostly the original topic.
Last edited:
John Sheehy
Forum Pro
What was the lowest ISO? That makes a difference. Many cameras that are not ISO-invariant bewtween 100 and 200 are mostly invariant above a certain point in the ISO scale. It also makes a difference how low the darkest tones are in the weakest RAW color channel.
I don't believe that any current camera that uses twice the gain for double the base ISO is truly invariant starting at base, if you allow all real noise to come through. The ISO 100 tends to have more chromatic noise, not visible in a conversion that says "no thanks" to fine chromatic details, or nothing is actually close to black in any color channel.
Any mathematical manipulation of RAW values also can suffer more at ISO 100 than 200, in the deep shadows.
John Sheehy
Forum Pro
A lot depends on how much the post-gain read noise character deviates from purely random. Let's take the Nikon D5, for example. It has very low pre-gain read noise, but it has one of the highest post-gain read noises of recent cameras, and it is heavily banded. The 6D2 has less post-gain read noise, and it is much more random, so ETTR at a higher ISO is less important with the 6D2 in the higher ISO range, even though it still makes a tiny difference. The D5 benefits highly from using ETTR at a higher ISO setting than normal for the ISO exposure index. That is true, even if the D5 has less noise than a 6D2 or D810 doing the same thing. There are absolutes, and there are relative compromises within a system.
The Ghost of Caravaggio
Senior Member
- Messages
- 1,006
- Solutions
- 1
- Reaction score
- 747
Cameras that license Aptina's dual-conversion gain technology essentially have two, psuedo ISO-invariant ranges.
Just one example are FUJIFILM XTrans III cameras. From ISO 200 - 640 the full-well capacity is optimized. From ISO 800 and above the photo-diode sensitivity is optimized.[1] Other brands use slightly different ISO ranges, but the result is the same – two pseudo ISO- invariant regions.
The sensor electronics actually change the conversion gain. I think of this as having two base ISOs. One for scenes with high dynamic range and another for scenes where the sensor's DR is not exceeded but the light level is low.
1/ I used the word sensitivity because that's what Aptina's engineers use. But what do they know?
Just one example are FUJIFILM XTrans III cameras. From ISO 200 - 640 the full-well capacity is optimized. From ISO 800 and above the photo-diode sensitivity is optimized.[1] Other brands use slightly different ISO ranges, but the result is the same – two pseudo ISO- invariant regions.
The sensor electronics actually change the conversion gain. I think of this as having two base ISOs. One for scenes with high dynamic range and another for scenes where the sensor's DR is not exceeded but the light level is low.
1/ I used the word sensitivity because that's what Aptina's engineers use. But what do they know?
dwalby
Veteran Member
OK, thanks for the clarification, I was trying to keep things simple but it would appear that was a mistake. I inconveniently forgot that some of the noise prior to the amplification is referred to as read noise, which makes a huge difference in the SNR going into the amplifier, so that was a big disservice to the OP. sorry about that.
yeah, I think I need to stay out of these types of discussion in the future.
Last edited:
JohnTheKeenAmateur
Well-known member
You have piqued my curiosity, John ... Can you pls provide some examples ?
(Pointed here by richarddd ... Thank you, Richard ... I'm lurking & learning)
bobn2
Forum Pro
The problem that many people have with this conversation is that they can't get out of their head the wrong idea that ISO is 'amplification'. It isn't, 'amplification' is irrelevant to the 'ISO', but it isn't necessarily irrelevant to the results that a camera produces when set to a particular ISO.
First to explain why 'ISO' isn't 'amplification' (apart from the fact that ISO says it is something else). What people tend to get wrong is that the input and the output of the photographic process is the same thing - light, they then think that 'light' has to be 'amplified' to get a photo from less light. In fact the input and output are different kinds of thing. The input is exposure, light energy density (generally measured in lux seconds) the output is a location (value) in a colour space, which is a perceptual specification, how bright and what colour that point appears to be. It only has an indirect connection to an amount of light is that how much light that 'value' generates depends on the display device.
So, to convert between different types of thing, you don't need 'amplification', you just need to make the conversion, according to its definition. The job of the sensor and its associated electronics is to measure the exposure in each pixel. The conversion to the output values is a computational task. The reason that 'amplification' or 'gain' is changed is to enable the camera to make more accurate exposure measurements in low light, not, in essence, to produce 'ISO'.
The notion of ISO invariance is that for most camera over a part of the ISO setting range, changing the ISO is not making the exposure readings any more accurate, but it is chopping off the highlight capability (due to a rather silly convention in raw file format) so it often makes sense to use a lower ISO setting and keep the highlights, given that many low light scenes have light sources within the image.
You have convinced me that digital ISO is not true ISO amplification. It is not even close from what you and others proved to me, with the 100 vs 3200 comparison, in our last thread. But it is not the common amateurs' fault that the so called "pros" at every photo blog and web site call it amplification.
Hell I think it is mentioned in the first digital photography book I ever bought. They have been drilling that concept in our head, and still do, since the first digital cameras.
I am glad you posted here bobn2. I wanted to state the above in the other thread; but it had already reached its limit.
--
Common sense is common knowledge; not everyone's common knowledge is the same.
Last edited:
bobn2
Forum Pro
JohnTheKeenAmateur
Well-known member
<snip>
I may have an understanding but now I have to figure out a best way for putting it in practice with low light photography.
Which is better for an under 30 second exposure?
ISO 100, f5.6 - F8.0
ISO 400, 5.6 - F8.0
ISO 800, F5.6- F8.0
ISO 1600, F.56 - F8.0
Or does the ISO even matter and just let the camera choose? My goal is simple; capture as much light as possible with the least noise. It is time for me start all over with my way of thinking in terms of ISO.
--
Common sense is common knowledge; not everyone's common knowledge is the same.
Last edited:
crashpc
Veteran Member
Depends...
You have to understand that the exposure is only made of f number and exposure time.
If you don´t have such time or aperture restrictions, that your exposure fills into the histogram easily, then shoot lowest ISO speed.
If you reached your maximum exposure, and the image would be still dark at ISO 100, you can do two things.
For "ISO-Less" like modern Sony, you can let it be and brighten the image in the postprocessing, if you shoot RAW. If you shoot JPEG, Increase ISO speed so the final image brightness is good for you.
For older sensors like Canons up-to-24 Mpx (not the one in SL2,80D,77D,M5,M6,760D and so on), always increase your ISO speed, so you get your image bright enough. But not too much to blow highlights, which gets easier with higher ISO speeds.
This technique has been mastered, and is called "BTTR" - Brightness To The Right (in your histogram).
You have to understand that the exposure is only made of f number and exposure time.
If you don´t have such time or aperture restrictions, that your exposure fills into the histogram easily, then shoot lowest ISO speed.
If you reached your maximum exposure, and the image would be still dark at ISO 100, you can do two things.
For "ISO-Less" like modern Sony, you can let it be and brighten the image in the postprocessing, if you shoot RAW. If you shoot JPEG, Increase ISO speed so the final image brightness is good for you.
For older sensors like Canons up-to-24 Mpx (not the one in SL2,80D,77D,M5,M6,760D and so on), always increase your ISO speed, so you get your image bright enough. But not too much to blow highlights, which gets easier with higher ISO speeds.
This technique has been mastered, and is called "BTTR" - Brightness To The Right (in your histogram).
Keyboard shortcuts
- f
- Forum