Bonsia ~ Testing Depth of Field

[Ysarex]

Your chart raises a few red flags. You watched the toothpaste video and you're showing possible influences from the Triangle Cult. Triangle Cult first: ISO does not control noise in a photograph and increasing ISO does not increase noise. Increasing ISO (especially since you're using a Canon camera) reduces noise. The triangulars are confused about cause and effect. They've attached themselves to a spurious correlation and converted that over to a cause - effect. Noise in a photograph is primarily a function of exposure. Less exposure = more noise and more exposure = less noise. ISO is related: When we don't have enough light to, as JC noted, achieve a fast enough shutter speed then we have to reduce exposure. If we don't have enough light to achieve the aperture we require for DOF then we have to reduce exposure. The noise is a result of reduced exposure. ISO does two things. It biases our camera meter to allow the reduced exposure calculation and then it post processes the sensor capture and brightens the image to compensate for the reduced exposure. In this post processing function ISO typically reduces the noise caused by the underexposure. It may sound like a subtle distinction but it's important to keep cause and effect straight. You can't control noise in a photo with the ISO setting; you have to do that with exposure. Noise is a function of exposure.

Joe

Joe,

I'm wondering if I've misunderstood what you meant to write.

Noise is a tricky subject to tackle and I honestly don't know anyone who feels completely comfortable tackling all aspects of it. One of the reasons is that noise is caused by lots of contributing factors... not just one thing.

Noise is more noticeable as a result of amplification ... which is the gain applied as a result of dialing up the ISO setting. (Changing ISO does not alter the sensitivity of the sensor.) This gain (amplification) is either applied via analog means prior to analog to digital conversion (ADC) and is sometimes called "upstream gain", or it could be applied via digital multiplication after ADC and is sometimes called "downstream gain", or a combination of both. How it's performed varies by camera manufacturer and model.

But noise is caused by lots of other factors... heat build up (which does tend to increase in long exposure images), even quantum effects (that we can't control). One type of noise is "read noise" and this is an amount of noise inherent in a sensor (due to electronics) and that amount of noise is considered to be fixed.

We regard an image as "not noisy" not because there is no noise... but because we don't notice it. This is a result of the "signal to noise ratio" (SNR). So if "read noise" is fixed... but I can get more "signal" by taking a longer image, then it stands to reason that I will have a better SNR by taking a longer exposure image than a shorter exposure image. This supports the hypothesis that a longer exposure duration should result in less noise (which is what I believe you are saying). But this is primarily a function of read noise. But there are other types of noise.

This is why I started with "I don't know anyone who feels completely comfortable tackling all aspects" of noise.

I can say that as an astrophotographer (where shooting really long exposures at high ISO is pretty common)... it doesn't matter that I can take a really long exposure. We still get more noise at higher ISO levels (regardless of exposure duration). (Astrophotographers deal with this by taking loads of images of the same object and then using "stacking" software to integrate the images ... especially using statistical integration algorithms which do a pretty impressive job of knocking down the noise.)

If the hypothesis that noise is primarily just a function of exposure duration, then we should be able to predict that if, in addition to increasing ISO, we also increase the exposure time, then we should not see an increase in noise because we've maintained the exposure duration.

So for example:

ISO 100 at f/2.8 for 30 seconds ... compared to ... ISO 12,800 at f/32 for 30 seconds
(trading 7 stops of aperture for 7 stops of ISO so we can maintain the same 30 second exposure duration).

But if I do this test... I get no noticeable noise in the ISO 100 image, but loads of noise in the ISO 12,800 image.

The data does not support the hypothesis that noise follows exposure time (instead of ISO).

I never suggested noise follows exposure time. Noise follows exposure. In your example the second shot taken at f/32 is exposed massively less than the first shot taken at f/2.8. And that's the cause of the noise. Exposure = quantity of light per unit area reaching the sensor: function of ambient light level + time + attenuation through the lens (f/stop).

I could also shoot at ISO 12,800 but trade 7 stops of aperture for exposure duration (so f/2.8 at 1/4 seconds instead of f/32 at 30 seconds). This also results in a noisy image. There may technically be less total noise in the longer exposure due to the improved SNR ratio w.r.t. the "read noise" component of the total noise, but because this is being overwhelmed by other contributors to noise, I don't notice an obvious improvement to the image. The only way I do notice a dramatic and obvious difference, is to lower the ISO value.

In testing that I've done... noise primarily follows ISO setting (at least with my cameras). I have specifically tested the scenario where I set a low ISO with a long image duration (30 seconds) to get a representative sample of how much noise I get. Then used a very high ISO but still use the same 30 second exposure time (compensating by stopping down aperture) and can confirm that the long exposure does not help smooth out the noise problem in any meaningful way. The only way to get a meaningful reduction in noise from a single image is to use a lower ISO value.

There are loads of contributing factors to noise. And one major wrench-in-the-works is that accurately assessing a sensor relies on having information available which manufacturers do not share. We do (now) know that many manufacturers "cook" their RAW images (in other words the true read noise & shot noise is higher than they let on... but they de-noise the images ... and still tell us these "cooked" images are "RAW") and this makes it difficult to accurately assess the camera at a technical level.

In science, if we observe a behavior and try to ascribe a cause, we might be able to do that if there is only one cause. If there are multiple causes, it is difficult to understand how much of the observed behavior should be ascribed to one factor vs. another factor. If we can gain confidence that we thoroughly understand one of those factors, then we can more accurately understand the other factor (because it's responsible for whatever is left over). But if you have multiple factors (more than two) ... especially when you don't know how many factors you're dealing with (and some of it's trade-secret) it gets really hard to be able to say we understand or can accurately explain the observed behavior.

So I wont attempt to say "why" noise follows ISO (more than exposure duration). I can only state that testing the camera shows that it does.

Tim

Most of us are not astrophotographers and the noise due to heat from long exposures isn't an issue for the overwhelming majority of photos taken - like +99% of them. We can set that aside as special case.

That leaves us with the two primary sources of noise: shot noise and read noise. Given today's modern sensors, read noise is reaching the point of negligibility compared with shot noise (think ISO invariance). When people today notice noise in their photos they're basically seeing shot noise. And the cause of that noise is reduced exposure -- failing to fully expose the sensor.

I simplified based on what people in fact are really experiencing.

The ISO brightening applied as either analog gain or digital scaling or both amplifies the noise along with the rest of the signal but, the critical point: it is not the cause of the noise. The cause is failing to adequately exposure the sensor.

Given the colloquial presentation of exposure controls and ISO there's a tendency to attribute causation for noise to ISO. Yes, they're linked via exposure. When folks raise the ISO value on a camera they're typically doing that to calculate an exposure reduction. Then they get a noisy photo and say ISO caused the noise. No, the exposure reduction caused the noise.

In cameras (most cameras) where analog gain is used as the method of ISO brightening the applied gain improves the result due to read noise and so applied ISO brightening in most cameras helps suppress noise.

Joe

 

[TCampbell]

I never suggested noise follows exposure time. Noise follows exposure. In your example the second shot taken at f/32 is exposed massively less than the first shot taken at f/2.8. And that's the cause of the noise. Exposure = quantity of light per unit area reaching the sensor: function of ambient light level + time + attenuation through the lens (f/stop).

Joe

Ahhh... that makes MUCH more sense. And yes, I agree.

I had previously thought you were linking it to exposure duration.

Another way of thinking about this: It's somewhat common to read that "ISO changes the camera's sensitivity to light". ISO doesn't actually change the sensitivity ... the chip is as sensitive as it is. However much light was captured during the exposure (based strictly on aperture and shutter time and ignoring ISO) is whatever it is. Boosting ISO is simply applying amplification to whatever came in... after the fact.

It's a bit like radio. If the radio station is close and the signal is strong, you get a nice clear signal. If the radio station is distant and the signal is weak, you might hear a lot of static. You could "amplify" the weak signal to make it louder, but that amplifies _everything_ (both the signal you want to hear... and the noise & static you don't want to hear ... it all gets louder).

If you can use an exposure that adequately captures the shot without boosting ISO, then that would be a much cleaner image.

 

[Ysarex]

Your chart raises a few red flags. You watched the toothpaste video and you're showing possible influences from the Triangle Cult.

Toothpaste: The toothpaste video isn't too bad but it doesn't simplify DOF; if anything it complicates it and confuses some important aspects.

Toothpaste didn't manage to work in the role of Coc (circle of confusion) in the graphics.

With the subject closer to the camera (increase magnification) the DOF decreases.

View attachment 152301

Joe

Well Joe, what can I say, we go from red flags, to Triangle Cults, to Circles of confusion and then include noise. I am glad you didn't put those in capitals. (Just joking).

I can see a little in what you say, but being a total beginner I need to hang on to what I learnt so far and put that into practice, practice, practice, before going deeper. That said I am always open to comments, that was why I posted the image, and even if I am maybe not ready for the extra depth of understanding yet, other readers may be.

For me at this stage, I have proved to myself and had it confirmed on here that as far as DoF is concerned in the image, which is all the test was for from my prospective, I have grasped the basics of controlling DoF. I now need to find other items and do things differently, probably testing other things, the more I learn by using the shutter speeds, aperture and ISO, one against the other the sooner I will get it to gel in the brain and then will become second nature, allowing the mind to concentrate on the items and light in the image, which way they come from, etc, etc. Sadly until the better weather I cannot get out, I cannot even hobble to the car when the wind is 20mph or more and at present gusts in the mid 60's are forecast. There is a nature reserve within 10 miles, I know my lens are not powerful enough for many of the birds at distance but it will be worth a visit when the weather gets better, there is plenty more items to shoot at not just the wildlife.

Apologies -- I tend to over do it at times.

In terms of DOF this may be still helpful and new info: The change in DOF when you change from one f/stop to the next is approx. 50%. The change in DOF when you change magnification by a factor of 2 is a factor of 4. In other words move a camera closer from 6 meters to 3 meters (factor 2) and the magnification will double while the DOF zone will shrink to 25% of its previous depth. Same thing works with focal length. Take a photo from 6 meters with a 50mm lens. Switch to a 100mm lens and the magnification will double while the DOF zone will shrink to 25% of its previous depth. Works in reverse the same way. What you then discover is that magnification has a greater effect on DOF than does f/stop.

Joe

 

[jcdeboever]

As promised

 

[Ysarex]

I never suggested noise follows exposure time. Noise follows exposure. In your example the second shot taken at f/32 is exposed massively less than the first shot taken at f/2.8. And that's the cause of the noise. Exposure = quantity of light per unit area reaching the sensor: function of ambient light level + time + attenuation through the lens (f/stop).

Joe

Ahhh... that makes MUCH more sense. And yes, I agree.

I had previously thought you were linking it to exposure duration.

Another way of thinking about this: It's somewhat common to read that "ISO changes the camera's sensitivity to light". ISO doesn't actually change the sensitivity ... the chip is as sensitive as it is. However much light was captured during the exposure (based strictly on aperture and shutter time and ignoring ISO) is whatever it is. Boosting ISO is simply applying amplification to whatever came in... after the fact.

Yeah, Don Quixote here. When I fully retire I can ignore the windmills but for now this continues to plague me. I'm starting a new semester now and here I go again. Every semester I get students in class I have to deprogram from the Triangle Cult. They get sucked in on Youtube before I get them. You're immune but, take an impressionable beginner and before you know it they believe ISO is an exposure determinant. Youtube tells them that each of the three vertices of the triangle balance against the other two. Therefore:
1/500 sec, f/8 and ISO 200
1/125 sec, f/16 and ISO 200
1/500 sec, f/11 and ISO 400
all balance as equivalent exposures. And ISO causes noise. PITA, so when I see a beginner here at TPF I sometimes try and save them before they get sucked in.

Joe

It's a bit like radio. If the radio station is close and the signal is strong, you get a nice clear signal. If the radio station is distant and the signal is weak, you might hear a lot of static. You could "amplify" the weak signal to make it louder, but that amplifies _everything_ (both the signal you want to hear... and the noise & static you don't want to hear ... it all gets louder).

If you can use an exposure that adequately captures the shot without boosting ISO, then that would be a much cleaner image.

 

[PJcam]

I can see a little in what you say, but being a total beginner I need to hang on to what I learnt so far and put that into practice, practice, practice, before going deeper. That said I am always open to comments, that was why I posted the image, and even if I am maybe not ready for the extra depth of understanding yet, other readers may be.

As a simple observation, it seems to me that you're making this way too complicated. I know you wrote that we each need to learn how we do it best, but learning DOF by experimentation seems the long hard way as compared with consulting a chart or website.

You learn your way, and I'll learn my way, but learning DOF wasn't exactly difficult for me, so I just accept it for what it is and move on.

No problem Designer, we are also all at different stages. For me, I have a camera and a couple of lens, purchased as a bundle, I know where the shutter button is and the lenses go on the front, I still haven't found out where to put the film! That puts where I am in prospective, I have had the camera 2 weeks, I have one foot on the learning ladder it is only on the first step.

Instead of getting to know how Shutter speed, Aperture and ISO work or how they need to be set to work together, I stepped ahead, I wanted to understand DoF, for me it just seems an important part of some or many finished pictures. So I did what made sense at the time, try in and see, practice and learn, I also did what the forum suggests, go on share your image, don't be shy. The forum encourages us all, no matter how much or how little we know, or don't now, to share and ask questions.

I honestly cannot wait for the better weather to be able to get out and experiment outdoors. Now where does that damn film go.

 

[PJcam]

Interesting discussion guys, but in some ways I cannot help but think, over 90% of people would not be concerned to the level of thought here, but are more concerned with reducing noise (or using) noise to get the best from their shot.

My next thought, as a total beginner, is although there are several (more than two) items influencing noise in the image, probably one of them is a bigger influence than the others. @TCampbell uses the word, 'amplified' this one seems to me, at this stage of understanding, to be a better description of the issue relating to noise.

I watched this YouTube video the other day regarding ISO and noise...


The explanation in the video made sense and shows clearly that amplification is a big part of the issue. (We have accepted there is more than one reason).

'Amplification' as shown in the video is relating to and causes noise, or increase noise, amplifies the noise. I am thinking maybe it is a little bit like taking a medium resolution image and altering it's size to make the image larger, doubling it's size, as a result we amplifying any noise already in the image so the quality of the image deteriorates.

I this wrong thinking?

 

[jcdeboever]

Keep it simple. ISO is a tool to gain shutter speed.

The above lesson and guide makes it real simple to understand DOF, that's what I love about John Hedgecoe books, simple with lots of pics. @Derrel recommended them to me about 2 years ago when I was first starting out. They're old but still apply. I paid like a $1 for almost everyone I own. I have literally done over 100 projects contained within them. It got me out shooting and associating with the text to expedite learning.

 

[Designer]

[/MEDIA]

The explanation in the video made sense and shows clearly that amplification is a big part of the issue. (We have accepted there is more than one reason).

'Amplification' as shown in the video is relating to and causes noise, or increase noise, amplifies the noise. I am thinking maybe it is a little bit like taking a medium resolution image and altering it's size to make the image larger, doubling it's size, as a result we amplifying any noise already in the image so the quality of the image deteriorates.

I this wrong thinking?

Wrong? I say more like mislead.

There are probably several videos available that purport to teach the technology of digital photography, and the problem is to figure out which ones are correct.

I started the video you referenced, and noticed that he made a fundamental mistake at only 18 seconds into the video. The sensitivity of your sensor doesn't actually change. It is what it is from the day of manufacture until it is finally scrapped and sent to the Great Recycling Bin in the sky. Stays the same. Does not change.

As @TCampbell wrote; ISO is applied gain. That kind of makes it a requirement that the data has already been collected. (exposure made, file saved) Then what part does the ISO setting play? Deep within the electronic components of your image processor (the camera) the electrons get to work and produce a viewable (to human eyes) image and display that image on the LCD.

The short version:

1. Some cameras handle applied gain better than others.

2. Some exposures test the hardware and firmware to a practical limit (underexposures or exposures in low light conditions).

If those factors are put to their limits, then you see artifacts of electronic noise.

 

[Ysarex]

Interesting discussion guys, but in some ways I cannot help but think, over 90% of people would not be concerned to the level of thought here, but are more concerned with reducing noise (or using) noise to get the best from their shot.

Absolutely and I think it's fine if most people just black box it and learn to use it to get repeatable results. I can even just stand there and smile when I hear someone say ISO adjusts the light sensitivity of your camera. Bleepin bleep Nikon even says that on their web page (don't read that!)-- talk about misinformation!

But I can't let a room full of photography students keep functioning that way. And I do see that the misunderstanding has real ramifications that show in their photos -- their confusion causes them to reach wrong conclusions and do the wrong thing. So OK I'm letting my other life spill over in The Photo Forum -- sorry there.

As a related aside to this I recently had to create an test example of my Fuji X-T2 used at ISO 25K -- that's an APS class sensor used with the ISO set to 25,600. I went down in the basement (very dark) and took a photo of my canning shelf. Must be gosh awful noisy: green beans

My next thought, as a total beginner, is although there are several (more than two) items influencing noise in the image, probably one of them is a bigger influence than the others. @TCampbell uses the word, 'amplified' this one seems to me, at this stage of understanding, to be a better description of the issue relating to noise.

I watched this YouTube video the other day regarding ISO and noise...


The explanation in the video made sense and shows clearly that amplification is a big part of the issue. (We have accepted there is more than one reason).

'Amplification' as shown in the video is relating to and causes noise, or increase noise, amplifies the noise. I am thinking maybe it is a little bit like taking a medium resolution image and altering it's size to make the image larger, doubling it's size, as a result we amplifying any noise already in the image so the quality of the image deteriorates.

I this wrong thinking?

Your analogy of enlarging a lower res image isn't bad but that video is very bad and very wrong. The Internet has effectively reduced the cost to publish to $0. The old method of publishing on paper wasn't fool proof but at least someone at some level who had to open their wallet and pay to run the press was entitled to ask the question, "what is it were publishing and is it worthwhile and has someone checked it over for mistakes?"

Amplification? yes and no. We have a case here of finding the right words to fit what's going on -- I like boost. As usual it tends to be a bit more complicated. One method employed in our cameras is analog signal amplification -- the electrical signal (photons to electrons) is boosted through an amplifier. The word amplification has connotations that don't entirely match up and more and more in our modern cameras, because the only source of noise that the video identifies is no longer a factor, we're foregoing the analog amp stage and replacing it with the much less expensive digital scaling (strictly numbers) stage and/or combining the two methods. There's no analog amp in the Nikon D7000 so it's tricky to say that the signal is amplified. My old X-E2 combined both methods.

The noise we see in our images today isn't coming from the camera circuitry (heads up Youtube). We've effectively beaten that down. Those were the good old days. The noise we see now in our photos is the noise contained in the signal itself. A weak signal is noisy and that's what is causing noise in our photos. That's why I said noise is exposure (and ISO does not determine exposure). Here's another fun fact: Digital camera sensors don't have assigned ISO values. The ISO values on your camera do not refer to the sensor in your camera -- not even the base ISO value. They reference the output from the camera's JPEG processor.

Joe

 

[Braineack]

The noise we see in our images today isn't coming from the camera circuitry ... The noise we see now in our photos is the noise contained in the signal itself.

how is the signal generated if not from an electrical circuit?

 

[Ysarex]

The noise we see in our images today isn't coming from the camera circuitry ... The noise we see now in our photos is the noise contained in the signal itself.

how is the signal generated if not from an electrical circuit?

Thanks -- precise language is always hard. Shot noise is in the incoming signal (photons) that reach the sensor -- the light. It's not added to signal leaving the sensor (electrons) by the circuitry. Read noise is noise from the circuitry and has been a real problem in earlier cameras. It's still there but it's suppressed now to a level where it's nominal and not the noise that we see in our low-light photos with modern cameras.

Joe

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

 

[Braineack]

that's pretty much what the video suggests...

 

[Ysarex]

that's pretty much what the video suggests...

The video doesn't mention or suggest shot noise at all. And so viewers are likely to draw the wrong conclusion as to the noise source. What's that noise? Part one: Shedding some light on the sources of noise

Joe

P.S. Went back and looked at the video more carefully and in fact he's confused. He talks about the noise be amped up along with the signal when ISO gain is applied. That's what happens to shot noise and upstream read noise, but the noise he describes as caused by the electronic circuitry is in fact downstream read noise which is not amped up along with the signal but applied after ISO gain.

 

[PJcam]

Ermmmm. I think, if I keep ISO as low as possible, get best shot I can, learn from it and constantly look to improve, I am probably on the right track. The key being I cannot get a good image from a bad one (as the example previously discussed) so get the best shot(sssss) I can.

Got to go I have a lot to learn.