Manually lowering the ISO speed (<100)

[aroy]

Low ISO generally gives less noise, and the lowest is the native ISO, 100 in most CMOS sensors. Any lower ISO will be achieved by attenuation of the signal from the sensor. Higher ISO is achieved by amplifying the signal. Many cameras provide the "Lo" range, most do not.

For starscapes you go for higher ISO rather than lower ISO. To lower the noise you can do image stacking, that is take a series of shots, say at1 sec interval and then using software add them up. This technique reduces the noise so that it is much less than in an individual frame. Why do it rather than expose for a long time? With digital you increase the noise as the exposure time increases, more so as the sensor heats up (or in hot weather), so you want the noise in each shot to be within a reasonable amount. By stacking a series of shots you are increasing the light (so that the starts are brighter) and at the same time reducing noise (random noise to be precise).

To reduce the quantum of light entering the camera, you use ND (Neutral Density) filters. These come in various "strengths" and reduce the light by 1EV to 10EV. Variable ND filters are generally two polarizing filters. as you change the angle of one with respect to the other the amount of light transmitted changes, so you can get an infinite gradations.

As is the case every where, good filters cost a lot of money. What do you pay for?
. Glass/polymer with uniform transmission across the field.
. Better materials for the filter holder
. In most cases multilayer surface coating to reduce ghosting and reflections
. In some higher priced filters harder glass (less chances of breaking) and anti smudge coating, which ensures that practically nothing sticks to the surface - easier to clean and maintain

 

[Tom Grove]

aroy... that was extremely comprehensive! Thanks for that!!! One question... Are you saying I should use an ND filter for night shooting at higher ISO's?... or are the ND filters for only during the day? ... I have been doing the latter. I have 2 ND8's and a polarizing filter.

 

[dr_eyehead]

Presumably though there is a point where if you lower the iso too much by attentuating the signal you start to reach the noise floor, so beyond that you are not lowering the noise?

Also, someone mentioned that a variable ND filter I looked at that employed two polarised ND filtered had an uneven density. How can that be since as you say the variability comes from simply adjusting the angle of the polarisers?

 

[Tom Grove]

I know the polarizing filter can be rotated to be darker or lighter... the my ND filters are not variable... I (sometimes) just stack them along with my polarizing filter and adjust the polarizing filter to where I need it in extreme brightness... But I don't always need to stack them. Works well for me.

 

[dr_eyehead]

One polarised filter will not produce a variable attenuation, you need two for that. Don't understand what you mean, unless you are stacking a variable filter on top of other filters, which then begs the question why? because a variable filter using two polarisers will give you variability down to the darkest setting surely.

 

[aroy]

aroy... that was extremely comprehensive! Thanks for that!!! One question... Are you saying I should use an ND filter for night shooting at higher ISO's?... or are the ND filters for only during the day? ... I have been doing the latter. I have 2 ND8's and a polarizing filter.

No. There is no reason to reduce the light at night further, there is very little light as it is.

You can stack ND filters, but at times if the stack is too thick, you will get vignetting. That is why it is better to buy ND filters of larger diameter and use an adapter to fit them to your lens.

 

[Steve B]

Presumably though there is a point where if you lower the iso too much by attentuating the signal you start to reach the noise floor, so beyond that you are not lowering the noise?

Also, someone mentioned that a variable ND filter I looked at that employed two polarised ND filtered had an uneven density. How can that be since as you say the variability comes from simply adjusting the angle of the polarisers?

Just go out and Google "variable ND filters uneven" or something like that. You will find a lot of articles about it. Usually happens with cheaper filters and wide angle lenses. Remember that a polarizing filter is affected by the angle of the light as well. Using polarizing filters on ultra wide angle lenses will usually give you uneven results.

 

[dr_eyehead]

Just go out and Google "variable ND filters uneven" or something like that. You will find a lot of articles about it. Usually happens with cheaper filters and wide angle lenses. Remember that a polarizing filter is affected by the angle of the light as well. Using polarizing filters on ultra wide angle lenses will usually give you uneven results.

Normal light is randomly polarised though isn't it? So if you use two polarised filters, each with an even density then the density should not change as you adjust them relative to each other, all you are doing is restricting the polarity.

Is there an issue with the angle at which light is coming in through the filter not varying linearly with the strength of the filter maybe? I googled and someone was complaining about dark spots in the corners.

 

[Steve B]

Some have "hot spots" some generate dark cross patterns, there may be other issues with some. Just make sure to read the reviews on them before you buy. I don't use them since I would rather use ND filters. I use HiTech filters in a Cokin holder so I don't have to buy filters for every filter size. It's just like any filter. If you buy cheap you may not get the results you expect. There are a lot of variables with how filters are designed and constructed. Some ND filters introduce color casts even though in theory they shouldn't.

 

[aroy]

Presumably though there is a point where if you lower the iso too much by attentuating the signal you start to reach the noise floor, so beyond that you are not lowering the noise?
.............

To understand what a lower ISO will do (and what the "LO" range will not, here is a simplified version of how a sensor works.
. The digital sensor works by accumulating photons. The number of photons is limited by its design. The maximum number of photons that a sensel can hold is called the full well capacity, and that defines the base ISO.
. There are various types of noise involved in the chain - random, thermal and electrical. For this post I will consider only the noise generated in the sensel.
. The Dynamic Range (DR) of a sensel is roughly the ratio of the signal at full well capacity to the noise.
. To increase the ISO, you just amplify the signal from the sensel, and that amplifies the noise also. That is why the best DR is at base ISO, as above that the signal decreases while the noise remains more or less same. If you decrease the signal to simulate lower ISO, you will decrease the noise generated in the sensel, but then other noise components generated further down the stream will increase.
. After the signal from the sensel is processed it is digitized in an A-D converter. The number of bits in the output is decided on the basis of the speed of conversion and the number of photons in full well. The larger the number of bits in an A-D converter the slower it gets. 16 bits means 64K discrete levels and 14 bits 16K discrete levels. If the full well capacity is less than the number of discrete levels then there is no use having so many levels. The noise (and DR) further reduce the number of levels.
. Initial sensors had only one A-D converter, so that the process was very slow, as it converted every bit serially. Modern cameras have one A-D converter per row/column as the design may be, hence are much faster.
. The total burst rate of a camera is dependent on
- exposure time, including shutter cycling
- Read time for all sensels
- A-D conversion
- Demosaicing
- Writing to memory
- clearing the sensor of its charge - initializing it to zero
Thus a fast A-D and a fast processor are required for fast burst rate. That is one of the reasons that very fast burst rates are limited to 16/18MP sensors. doubling the burst rate from 5fps to 10fps will require the whole chain to be twice as fast. Similarly expanding the sensor from 16MP to 24MP will need 50% faster and from 16MP to 36MP 2.25 faster.

Lower ISO.
To lower the base ISO you have to redesign the sensor for lower ISO (larger full well capacity, etc), which will mean lower MP. Now unless you can up the DR proportionately, there will be no takers for lower ISO. Majority of the DSLR users are either sports, event shooters & press (high burst rate) or cost conscious users (lower the cost the better) or advanced amateurs (high end equipment with general interest). A lower ISO with enhanced DR will not appeal to any.

Lower ISO high DR is available with Medium Format, but the costs are 10-40 times that of the current 35mm DSLR, used by those who want excellent DR with great tonality as well as modularity - fashion & advertising requiring large product images and landscape photographers who want to print large. Because of better DR the tonal gradation in medium format is better compared to 35mm.

 

[Tom Grove]

One polarised filter will not produce a variable attenuation, you need two for that. Don't understand what you mean, unless you are stacking a variable filter on top of other filters, which then begs the question why? because a variable filter using two polarisers will give you variability down to the darkest setting surely.

I'm stacking 2 ND 8 filters with one variable polarized filter... I do that when I'm taking pictures and the sun is facing me. ( I realize it's not an optimal thing to take pictures when facing the sun... but sometimes I know I will not be there again when lighting is optimal).