High ISO Questions

[table1349]

If this doesn't clear it up, nothing will.

Noise, Dynamic Range and Bit Depth in Digital SLRs

 

[Ysarex]

If this doesn't clear it up, nothing will.

Noise, Dynamic Range and Bit Depth in Digital SLRs

I hope you have an industrial size popcorn maker -- get started.

Joe

 

[smoke665]

@Ysarex Not as up on this as you guys, but part of my understanding of the Higher ISOs was that they included an upgrade to the sensitivity of the sensor. As a lot of the downstream noise is generated by variance in the light being captured, wouldn't shortened exposure times decrease the amount of shot noise, but the DR remain the same due to the increased sensitivity?

 

[Ysarex]

@Ysarex Not as up on this as you guys, but part of my understanding of the Higher ISOs was that they included an upgrade to the sensitivity of the sensor. As a lot of the downstream noise is generated by variance in the light being captured, wouldn't shortened exposure times decrease the amount of shot noise, but the DR remain the same due to the increased sensitivity?

Shot noise is the primary source of noise. Shot noise is exposure period. More exposure = less noise and vice versa. It really is just that simple. Noise from the camera electronics by comparison is minor. It's not the light sensitivity of the sensors that's being increased as much as their total recording capacity -- DR range. I guess you can think of that as a sensitivity increase in the low end, but I think the key is that as we increase the DR range of the sensor we push the shot noise farther and farther away from a base ISO exposure -- all ISO exposures benefit.

Secondarily now with the D500 we get further improvement in the processing of the sensor signal to squeeze out every last drop by using Sony's new dual gain sensor.

Take a journey through time along with a re-read of Derrel's post from earlier today: Photographic Dynamic Range versus ISO Setting

Bill Claff is a great independent source for data. That tech article that Gryph linked is explained on Bill's site. I listed on the linked graph Bill's data for:

Nikon D100
Nikon D90
Nikon D7200
Nikon D750
Nikon D500

Note the increase in DR capacity over time. Each time the line goes higher the sensor's DR capacity is increasing.

Each time we increase the sensor's DR capacity we push the noise farther and farther away. With the D100 the noise was only 3 stops away from a base ISO exposure. With the D7200 that noise is 6 stops away from a base ISO exposure. And with the D750 it's a solid 7 stops away. That's a staggering breathtaking increase over 12 years time. Nothing of that magnitude ever happened in the progress of film tech.

Now look at the graph for the D500 (orange). It goes farther down than the other cameras because the ISO is adjustable to higher values. Compare it with the D750 at ISO 51K and the D750 has a higher DR rating. Why take the D500 so far then if it's not keeping up with the older D750? Because the increase on the other end from Sony's sensor is doing a better job at noise processing so why not. But Bill lists only a 2.16 DR for the D500 at 51K. This is where the complaints about these 51K and higher ISO values have some merit. Once you drop below 4 stops DR range on the sensor you're running a deficit overall that has to be made up with some heavy lifting in post processing. Drop below 2 stops DR on the sensor and you should start feeling like Micky here:

You don't have enough DR left to fill 1/3 of a normal histogram.

Joe

 

[astroNikon]

..

A lot of shooters hold ideas that are four,three,two camera generations behind what is possible with TODAY'S sensors, camera's and their internal electronics, and today's software

Think you nailed It!

you can research all you want but until you make the jump you'll never really find out.

I know many ppl would say a DX is close to a FX camera. But when I made the jump to a D600 from a d7000 I was totally amazed by the low light ability. So now I have the d500 and d750. but then I tend to shot in more extreme lighting conditions.

If you compare a D700 12mp FF to a D6x0 24mp FF, you'll first notice that the d700's pixels are, for simplicity, twice the size of the d600s. But the D600 of course is much better in low light. And a change from Expeed to Expeed 3 processors (wiki has a large page about this too).
So it's an entire system that is moving forward not just the sensor.

 

[smoke665]

@astroNikon my research was more in understanding what I already have. The K3ii has a max ISO of 51200, but frankly it's not that usable. The new KP with 819,200 Max brings the bar up on the low end, making 51200 usable but sacrifices some of the things I like about the K3ii. Supposedly there's one in the works to replace the K3ii that will stretch the ISO limit higher yet, so that's the one I'm waiting on.

 

[smoke665]

More exposure = less noise and vice versa. It really is just that simple.

Don't you mean "LESS" Exposure time = less noise. As in the less time the shutter is open the less chance there is for light variations to hit the sensor????

 

[Ysarex]

More exposure = less noise and vice versa. It really is just that simple.

Don't you mean "LESS" Exposure time = less noise. As in the less time the shutter is open the less chance there is for light variations to hit the sensor????

Shot noise is the dominant noise source and shot noise decreases with more exposure and increases with less exposure: What's that noise? Part one: Shedding some light on the sources of noise

Joe

 

[astroNikon]

@astroNikon my research was more in understanding what I already have. The K3ii has a max ISO of 51200, but frankly it's not that usable. The new KP with 819,200 Max brings the bar up on the low end, making 51200 usable but sacrifices some of the things I like about the K3ii. Supposedly there's one in the works to replace the K3ii that will stretch the ISO limit higher yet, so that's the one I'm waiting on.

It all comes down to what do *you* consider acceptable based upon the type of photography that *you* do.

Do you need fast shutter speed, shallow DOF and high ISOs
or slower shutter speeds, deep DOF and high ISOs?
or fast shutter, deep DOF and high ISOs?

I went to BestBuy probably 4 times taking shots comparing the D500 to various cameras.
In all my tests and doing the exposure math I came to the conclusion that it would *not* be able to do the indoor sports scenario that I shot in, but it would be fine in other situations. Looking on paper it should have done just fine.

So I bought it (twice actually). And when I tested it my math proved correct. It could not.
But that doesn't mean it's a bad camera, just not able to do what my d600 and d750 can do in low & bad light.

My D7000 (from several years ago) was only good up to 1600 ISO, based on my preferences. But it went higher.

For instance, the D500 can go to 1Million ISO.
At 1/8000, f/32 at 1M ISO do you find this image acceptable ?
D500_ISO_Tests (28 of 28)
..
..
51200

..
..
25600

or this
other at 25600

..

btw, the last one was the D5500 at 25600

 

[Derrel]

More exposure = less noise and vice versa. It really is just that simple.

Don't you mean "LESS" Exposure time = less noise. As in the less time the shutter is open the less chance there is for light variations to hit the sensor????

Not to speak for Ysarex, but I think you are thinking about long exposure noise, and say the associated Long Exposure Noise Reduction setting that most d-slr manufacturers offer as an in-camera setting, which is something that was a big problem at one time. Pretty sure what Ysarex meant was that "more exposure" means a longer exposure time, rather than a faster speed, or a bigger lens opening, rather than a smaller one, and in general "more exposure" means an exposure that is at least somewhat close to what "the light meter says is the needed level of exposure".

 

[droaingsong]

@astroNikon my research was more in understanding what I already have. The K3ii has a max ISO of 51200, but frankly it's not that usable. The new KP with 819,200 Max brings the bar up on the low end, making 51200 usable but sacrifices some of the things I like about the K3ii. Supposedly there's one in the works to replace the K3ii that will stretch the ISO limit higher yet, so that's the one I'm waiting on.

It all comes down to what do *you* consider acceptable based upon the type of photography that *you* do.

Do you need fast shutter speed, shallow DOF and high ISOs
or slower shutter speeds, deep DOF and high ISOs?
or fast shutter, deep DOF and high ISOs?

I went to BestBuy probably 4 times taking shots comparing the D500 to various cameras.
In all my tests and doing the exposure math I came to the conclusion that it would *not* be able to do the indoor sports scenario that I shot in, but it would be fine in other situations. Looking on paper it should have done just fine.

So I bought it (twice actually). And when I tested it my math proved correct. It could not.
But that doesn't mean it's a bad camera, just not able to do what my d600 and d750 can do in low & bad light.

My D7000 (from several years ago) was only good up to 1600 ISO, based on my preferences. But it went higher.

For instance, the D500 can go to 1Million ISO.
At 1/8000, f/32 at 1M ISO do you find this image acceptable ?
D500_ISO_Tests (28 of 28)
..
..
51200

..
..
25600

or this
other at 25600

..

btw, the last one was the D5500 at 25600

Very well explained. At the end, it all boils down to what you like to do with it.

 

[smoke665]

It all comes down to what do *you* consider acceptable based upon the type of photography that *you* do.

To me I consider color or chromatic noise objectionable, shot noise - not so much.

I think you are thinking about long exposure noise, and say the associated Long Exposure Noise Reduction setting that most d-slr manufacturers offer as an in-camera setting,

Negative. Referring to the length of time the shutter is open. The longer it's open the more chance for random variations of light to strike the sensor.

 

[john.margetts]

I think you are thinking about long exposure noise, and say the associated Long Exposure Noise Reduction setting that most d-slr manufacturers offer as an in-camera setting,

Negative. Referring to the length of time the shutter is open. The longer it's open the more chance for random variations of light to strike the sensor.

Noise is not caused by random variations in the light hitting the sensor. Such variations are so small compared to the general level of light hitting the sensor they will remain invisible.

 

[smoke665]

I think you are thinking about long exposure noise, and say the associated Long Exposure Noise Reduction setting that most d-slr manufacturers offer as an in-camera setting,

Negative. Referring to the length of time the shutter is open. The longer it's open the more chance for random variations of light to strike the sensor.

Noise is not caused by random variations in the light hitting the sensor. Such variations are so small compared to the general level of light hitting the sensor they will remain invisible.

From: What's that noise? Part one: Shedding some light on the sources of noise

Probably the most significant and certainly least recognized source of noise is what we call 'shot noise' or 'photon shot noise'. In the simplest terms, this is you being able to see the impact of light's inherent randomness.

You might remember talk of photons from science classes, and the key thing to remember is that, although we perceive light as being pretty uniform, it actually travels as a series of packets. That is: light is quantized. These packets (photons) arrive at your eye or your camera sensor at random intervals. Because of the way the eye and the brain work, you don't notice this, but when you look at a scene, you're being bombarded by little packets of randomly occurring light from every part of that scene

Other sites have similar discussions on the randomness of light.

 

[TCampbell]

This depends on the sensor. Some sensors don't do any (or at least not much) "upstream" amplification and rely entirely on "downstream" amplification. Downstream amplification results in a loss of DR (1 stop of DR is lost for every 1 stop of ISO boost).

If the sensor does "upstream" amplification, then that happens before the analog to digital conversion (ADC) and you can get some ISO gain without much of a loss in DR.

This has me a little concerned. DR is a function of exposure period. The engineering implementation in the camera that processes the sensor signal can employ a boost to the analog signal prior to ADC, digital scaling in the ADC and/or a hybrid combination of both. Regardless of implementation the method used to process the signal will do a better or worse job of retaining the sensor data but not create data. Exposure creates data. The light sensitivity of the sensor is in no way altered by the methodology engineered into the system to process the sensor output. A full sensor exposure = maximum DR. Any reduction in exposure = reduction in DR.

Joe

Pass the popcorn.

Joe, part of this is correct... in that the "light sensitivity of the sensor" isn't altered. However... the dynamic range of the image saved to your memory card can be altered by boosting ISO. But this happens as a result of what happens to the data after it leaves the sensor. You certainly can lose DR by increasing ISO (there's lots of data to back this up.)

This is because you're dealing with "digital" data and you cannot record any values that overflow the bit-depth of the system. If we increase ISO by 1 stop then with "downstream" amplification it means we multiply all values by 2.

A 14-bit register can hold values from 0 to 16,383 (16384 possible values - that's 2^14th). So if we had a pixel with a value of, say, 10,000 (which is in the range that the camera can handle) and we double it.. it's now 20,000 and that's beyond what our 14-bit system can store. Data is clipped and we lose dynamic range. We can also try to "scale" the image (squeeze the histogram proportionally) to avoid clipping... (so really you're boosting shadows but not boosting highlights) but again... that means we're reducing the dynamic range. Either way, you lose dynamic range (but you can create an illusion of dynamic range because the data isn't clipping). (I suppose Nikon D-Lighting or Canon Highlight Tone Priority modes might be examples of this.)

If you're doubling the data, then you have a math limitation based on the bit-depth of the system.

If you boost exposure on your computer... instead of your camera... the computer has significantly higher bit-depth. I suppose one might say this is a bit of a cheat because you wouldn't boost ISO in the camera... you'd do in the computer (so the camera is shooting at low ISO where it still retains it's maximum dynamic range.)

I'm wondering what you mean by "DR is a function of exposure period." Can you elaborate on what you mean by that?

BTW, you can see plots of the data at dslr-astrophotography.com. Here for example, are plots for Nikon cameras.

Best ISO values for Nikon cameras

Take a look, for example, at the plot of the D3100 vs. the D3200.

On the D3100, you can see the DR remains reasonably stead as ISO increases from 100 to 200 and then 400. But when the ISO goes to 800 there's a slight (but not quite linear) loss in DR. When it increases to 1600 there's a steeper, but still not quite linear loss in DR. But from 1600 on there's literally a linear loss in DR for every stop of ISO.

Meanwhile on the D3200 there is immediately a linear loss in DR for every stop of ISO.

You can see that the D3200 has just over 12 stops of DR at ISO 100. The D3100 has just over 10 stops of DR at ISO 100.

But since the D3100 can make it to just about ISO 1600 before it starts the linear loss of DR, you can shoot a D3100 at ISO 1600 and get nearly 10 stops (I think they report that it's 9.6) of DR. If you tried to shoot a D3200 at that ISO, you'd get just a little over 8 stops of DR (it's hard to see based on the graph... maybe 8.5) So a D3100 outperforms the DR of a D3200 at ISO 1600 by a little more than a stop (maybe 1.5 stops) even though the D3200 outperforms the D3100 at ISO 100.

BTW, the bend in the graph of the D3100's ISO vs. DR plot represents the difference between "upstream" and "downstream" boost. When the graph is staying reasonably flat (at the left side) that's "upstream" boost. When the graph starts to bend downward (and ultimately goes linear) thats "downstream" boost. The manufacturers don't publish enough specs to tell us if and how much "upstream" vs. "downstream" they do, but you can reverse-engineer your way into it and figure out where it's happening for any particular camera model.