[jsherman999]

It is an interesting topic but also one that makes me realize why I like film much more. With digital we pixel peep the living crap out of a photo (a 16 MP image at web resoluiton is what 68 " long - who would look at a 68" long image from 10-12" inches away?) and scrutinize the smallest details which in the real world matter to very few people. It becomes too easy to scrutinize the technical and dismiss a photo without seeing the aesthetic qualities of it. I know I am guilty of it when I look at photos and expecially with my own photos. I probably trash too many because they are not sharp enough, too much noise, etc... when they might be a quite acceptable photo, just not at 1:1.

In my home office I have the Jurgen Vollmer shot of John Lennon standing in a doorway in Hamburg in 1961. It's filled with grain, very low-res, no DR to speak of... and fantastic, one of my all-time favorite R&R photos.

If we all were good enough (or lucky enough) photographers to have our images become immortal in the same way, shooting certain subjects, the technical issues would almost never matter. But we're not always shooting rock stars in a street - grainy, low-res motion-blurred shots of our loved ones in a living room or concert hall are not met with the same level of admiration

[dracluff]

That is a pretty fantastic shot. I enjoy how static he is compared to the movement of those in front of him.

I would say that to me most of the photos I have of loved ones are more records of time that I can go back to. I do try to get good ones every now and again that I will want to make larger but rarely do those happen in my living room. My children leave way to much clutter to make that possible. But if I ask my wife which ones she likes the best some of the times they are those poorer shots because they capture the personality her 'rock stars'. The photos most people shoot do not have to have the admiration of many, just the admiration of those people to whom the subject matters.

It might be what I like about film is that I give it more leeway to have faults and do not expect as much perfection from it. I also find it more difficult to get 'perfection' from it (because I can not just re-shoot after chimping). As a result of the lower expectations and the feeling of my hard work going into it I am typically more pleased (and get greater pleasure) when the results are good compared to my digital images. I do not mean to harp too much on digital nor start a which is better discussion because it is a personal taste and mine are no better than anyone else's. I think that digital is a great medium and has allowed for some really cool things as far as creativity goes. At the moment (and these views are subject to change at any given time) I find digital too sterile, too clean and well 'perfect' for me. I find more and more they do not represent reality as viewed through our senses but reality as viewed by a machine that can not err. But this can also be the photos I am looking at.
Anyways I will cut my off topic ramblings there.

1. Total Light. Since it has been stated quite emphatically (even by DPR) that Total Light has significant impact on IQ, can someone explain how a given fixed* quantity of light can affect the "non-cropped area" of an FF sensor? (I'm not elaborating here because I believe readers will know what I'm asking). If this Total Light argument is correct, then there must be a good explanation, even if it would take a physicist to explain.
Common sense, of course says that this Total Light concept cannot be right. Therefore...

The "total light" thing is right, only most people don't see the full picture. And that means that the important thing is the lens which gathers the light. Sensor size is practically irrelevant, if you have the right lens (fast and large image circle). That is why you can replicate any of todays FF DSLR images (incl. DoF, blur/bokeh and noise) on cheap APSC mirrorless cameras once you put a cheap lens turbo lens behind the FF lens. The lens turbo simply ensures all the light ends up on the sensor - the same total amout of light, even while the sensor has only 1/2,25 of the area. But that small piece gets 2,25 times as much light per area. It simply gets concentrated.
Thus you can laugh about the people stating there is no xyz lens (e.g. 24/1.4) equivalent on APSC, because there always is regading DSLR lenses: the same lens attached to a lens Turbo gives you an APSC 16mm F0.9.

2. Why, for any given level of technology, do FF sensors consistently show better high ISO performance (i.e. less noise)?

As equivalence dictates, that is never true. Comparable (equivalent) pictures always show the same noise. Only once you begin comparing apples to oranges, you gain differences. Even in real life you gain nothing from a large sensor once your subject requires you to leave the one extreme end of apertures. If you have to use say F4 on a faster lens since you say want to gain sharp faces at the same time from three familiy members on your Pictures, sensor size does not yield any differences, since the "equal DoF" requirement of your picture forces you to obey equivalence.

Be wary of little people who advocate equivalence-fundamentalism and then suddenly want to ignore it in other parts of a discussion where they dont like it any longer.
As stated above use a cheap APSC NEX with a lens turbo and even the part where people start saying "but you could not get the orange" goes away.

In another thread, Bossa suggested that this is merely a result of greater magnification applied to the APS-C image when performing the comparisons. This makes sense to me, and yet nobody has commented on this suggestion, the inevitable inference being that it has no merit.

It has more merit than 99% of posts on these types of topic. He's a smart man. Basically you do not need to understand a lot more than this piece. You never need to care about photons or anything like this here.
Magnification is the key item which changes viewers' perception of DoF, blur and noise.
As a fixed area section of a sensor of any size (large format or any crop size smaller than large Format like 6x7, 645, FF, APSC, mFT...) of same tech level always shows the same amount of noise "blotches" under comparable exposure.
Enlarge the picture a lot and the blotches get larger and more visible, too. --> more perceived noise.
Enlarge the picture and suddenly you can discern details you could not before --> you see more things "sharp" --> more perceived DoF, since DoF describes the range of what you can see sharp.

This also leads to the fact that once you crop a Picture in Lightroom (because you have such a nice high-megapixel-camera) you always increase noise and DoF. People who constantly crop photos in postprocessing could as well buy slower lenses and frame the origonal picture with more skill instead to achieve the same result.

[jsherman999]

The "total light" thing is right, only most people don't see the full picture. And that means that the important thing is the lens which gathers the light. Sensor size is practically irrelevant, if you have the right lens (fast and large image circle).

True.

That is why you can replicate any of todays FF DSLR images (incl. DoF, blur/bokeh and noise) on cheap APSC mirrorless cameras once you put a cheap lens turbo lens behind the FF lens. The lens turbo simply ensures all the light ends up on the sensor - the same total amout of light, even while the sensor has only 1/2,25 of the area. But that small piece gets 2,25 times as much light per area. It simply gets concentrated.
Thus you can laugh about the people stating there is no xyz lens (e.g. 24/1.4) equivalent on APSC, because there always is regading DSLR lenses: the same lens attached to a lens Turbo gives you an APSC 16mm F0.9.

Very true, and if you'e willing to relay on lens boosters that take away any hope of AF and accept an edge performance hit, you can simulate a larger look.

As equivalence dictates, that is never true. Comparable (equivalent) pictures always show the same noise. Only once you begin comparing apples to oranges, you gain differences. Even in real life you gain nothing from a large sensor once your subject requires you to leave the one extreme end of apertures. If you have to use say F4 on a faster lens since you say want to gain sharp faces at the same time from three familiy members on your Pictures, sensor size does not yield any differences, since the "equal DoF" requirement of your picture forces you to obey equivalence.

Well, there;s the magnification issue you seem willing to embrace but also, if you absolutely need to simulate aps-c or m43 DOF for some reason, when you stop down the FF shot you often get another benefit - more sharpness, less CA.

But the real power here is having the option that a smaller format can't provide - and that a larger format can (usually) because its a superset of that smaller format. You can (almost) always get a shot with FF that you can with aps-c - can't always go the other way.

It's also pretty powerful with the f/2.8 constant zooms - and the reason the Sigma 18-35 1.8 is so popular and exists in the first place. I don't know many folks who want to use a lens turbo booster on a zoom.

Magnification is the key item which changes viewers' perception of DoF, blur and noise.

Magnification or resampling is often the mechanism that makes it apparent, but total light is the reason the image SNR has increased in the first place. For example, take a D800 shot with the same** total light as a D7000, resample to D7000 (by cropping and displaying at same size,) and the images will look roughly the same. Take the shot with more*** Total Light, you will see the increase after resampling.

** Same total light for example would be: FF: 50mm f/4.2 1/100s vs. aps-c: 33mm f/2.8 1/100s aps-c.
*** More total light would be delivered at the same exposure: FF 50mm f/2.8 1/100s vs aps-c: 33mm f/2.8 1/100s

Be wary of little people who advocate equivalence-fundamentalism and then suddenly want to ignore it in other parts of a discussion where they dont like it any longer.

"little people" Don't worry, there's probably a lens turbo booster for you!

.

---

Last edited by jsherman999; 10-24-2014 at 08:30 AM.

Very true, and if you'e willing to relay on lens boosters that take away any hope of AF and accept an edge performance hit, you can simulate a larger look.

Well, we are mostly talking about the theories behind images here and there is no reason at all, why a lens turbo type adapter can not have 100% perfect AF. That is not an intrinsic advantage of a sensor size.
Then you need to see that - at least as far as photo forums go - Sony made the A7 thingies so very good. And most of the people saying so actually are adapting lenses due to the lack of FE ones. Thus there seems to be quite a bunch of forum users who are willing to go manual anyway.
And: As long as the total resolution actually is better on APSC than on FF probably many if not most will an edge performance hit.

Well, there;s the magnification issue you seem willing to embrace but also, if you absolutely need to simulate aps-c or m43 DOF for some reason, when you stop down the FF shot you often get another benefit - more sharpness, less CA.

That has as many downsides as upsides. If you use a fast lens the "more sharpness" oftentimes easily turns into "less shaprness" as fast lenses often deteriorate resolutionswise beyond f/4. And if you are shooting anything with a larger DoF requirement that is not what you like. I'd call it at draw.

But the real power here is having the option that a smaller format can't provide - and that a larger format can (usually) because its a superset of that smaller format. You can (almost) always get a shot with FF that you can with aps-c - can't always go the other way.

It actually is the other way around:

You can always get a shot on the smaller format using the wide angle adapter that you can get using FF. What you can not do most of the times is do it the other way around.
One disadvantage for FF you can not solve is that extenders always take a huge bad hit on Resolution, just as wide angle converter can improve it.
The other restriction is that FF usually is far behind in sensor Resolution so "cropping" in post is no option. Even today a FF would need to have 63 MPx to get there. That is not even on the horizon.
The third disadvantage with FF is that you never get a good optical viewfinder for the tele uses, because FF viewfinders are tiny compared to APSC in relation to the sensor sizes, thus the cropped frame in there is smaller than even in the cheapest entry elvel DSLRs today.
Not to speak of the disadvantage of size, weigth, price.

It's also pretty powerful with the f/2.8 constant zooms - and the reason the Sigma 18-35 1.8 is so popular and exists in the first place. I don't know many folks who want to use a lens turbo booster on a zoom.

Well again that is not something to argue about, it's just a personal taste thing. The number of People willing to drag around DSLRs is shrinking, so you could easily say that's a much larger Advantage of iphones versus FF DSLRs.
Much the same I could argue that FF has the large disadvantage of lacking 5 axis Image Stabilisation, which is quite popular and you can not achieve with a FF DSLR. Only that this is equally not a dierct consequence of sensor size as unwillinges to use converters with zooms is.

Magnification or resampling is often the mechanism that makes it apparent, but total light is the reason the image SNR has increased in the first place.

What is the hen and what is the egg?
For photography anything which "makes something apparent" on a picture is at the heart of things. If you touch stuff like SNR you take a Close look at the sensors themselves and there you find that the noise on the sensor for a given area is always the same (under comparable situations). This Impression only changes when you take a step back to look at the whole Thing and this already turns it into a matter of perception.


Just to make it clear: I find all sensor sizes have their value and I'd love to see affordable large format digital cameras. All crop formats you can buy today from 645 via 36x24 via APSC via mFT offer some good compromises depending on your priorities.


There is no "better" there is just a "more to my liking".

[ElJamoquio]

You can always get a shot on the smaller format using the wide angle adapter that you can get using FF. What you can not do most of the times is do it the other way around.

Huh? I don't agree, even if you ignore teleconverters.

[jsherman999]

That has as many downsides as upsides. If you use a fast lens the "more sharpness" oftentimes easily turns into "less shaprness" as fast lenses often deteriorate resolutionswise beyond f/4. And if you are shooting anything with a larger DoF requirement that is not what you like. I'd call it at draw.

In my experience (and with respect to my 'type' of shooting) it isn't close to a draw, because I'm never stopping down past a len's sweet spot just to get more DOF. When I do stop down for DOF, it's usually from f/1.8 to f/2.8 or f/4, and all of my lenses are hitting their sweet spot between f/4 and f/8.

This situation (stopping down for more DOF) doesn't apply to landscapes, etc for me because i"m almost always shooting them at about f/8 and taking advantage of hyperfocal, which I do both on aps-c and FF.

It actually is the other way around:

You can always get a shot on the smaller format using the wide angle adapter that you can get using FF. What you can not do most of the times is do it the other way around.
One disadvantage for FF you can not solve is that extenders always take a huge bad hit on Resolution, just as wide angle converter can improve it.
The other restriction is that FF usually is far behind in sensor Resolution so "cropping" in post is no option. Even today a FF would need to have 63 MPx to get there. That is not even on the horizon.
The third disadvantage with FF is that you never get a good optical viewfinder for the tele uses, because FF viewfinders are tiny compared to APSC in relation to the sensor sizes, thus the cropped frame in there is smaller than even in the cheapest entry elvel DSLRs today.

I find this ^^ a real stretch personally... It depends completely on using a turbo speed booster on aps-c one the one hand and discounts the ability to crop the high-MP FF mage on the other to get the same "reach", or to simply use a longer focal length. I think what you describe is a very labored 'advantage' for aps-c with those particular constraints.

What is the hen and what is the egg?
For photography anything which "makes something apparent" on a picture is at the heart of things.

Well, you were stating that it's the resampling that does the magic, I was pointing out that the 'magic' comes from the greater total light, the re-sampling is just a necessary mechanism in the process that shows it to you.

Saying that the advantage comes completely from the resampling is like saying that your car goes faster simply because you moved the gearshift from 4th gear to 5th gear, and that's the end of the story. It's what that gearshift engages that moves your car faster - the gearshift is just the mechanism, not the reason.

The egg doesn't hatch the hen that laid it

There is no "better" there is just a "more to my liking".

yep

---

Last edited by jsherman999; 10-24-2014 at 07:38 PM.

Huh? I don't agree, even if you ignore teleconverters.

That is a little too little to discuss. What significant "more" is there to be had on a 36x24 crop sensor compared to a 24x15 crop sensor?
Just look at the poor guys who are stuck with a Nikon FF DSLR. They can't achieve what the NEX APSC owner with a Canon 85/1,2 on a wide angle converter can. "More DoF control".
The Canon FF DSLR owners don't get the dynamic range the NEX owner gets.
None get the resolving power a NEX user with EF50/1,2 and WAC gets.

In my experience (and with respect to my 'type' of shooting) it isn't close to a draw, because I'm never stopping down past a len's sweet spot just to get more DOF.

That is a fair personal preference. Others have different ones.

I find this ^^ a real stretch personally... It depends completely on using a turbo speed booster on aps-c one the one hand and discounts the ability to crop the high-MP FF mage on the other to get the same "reach", or to simply use a longer focal length. I think what you describe is a very labored 'advantage' for aps-c with those particular constraints.

Well, it's a fun theoretical discussion anyways, so we all will find certain things a stretch. Always focussing on the one extreme end of lens uses (wide open) plus then assuming all others want this is at least the same stretch.
Using a lens Turbo lens is no more specific than talking about any other specific lens or camera. If you see the WAC (wide angle converter) as fixed attachment to the camera there is no "Stretch" left.
And cropping out APSC equivalent 24 Mpx or 28 Mpx from a 36x24 camera is simply impossible today. Much more than attaching an real life $150 WAC to any existing DSLR lens.

Well, you were stating that it's the resampling that does the magic, I was pointing out that the 'magic' comes from the greater total light, the re-sampling is just a necessary mechanism in the process that shows it to you.

I am fine with whatever anybody chooses to prefer here. :-)
You can say it's the fire that burns your skin while I say it's the heat from the fire. In talking to people who just take nice pictures and dont care about the theory behind it I found the explanation path using magnification a lot more compelling than anything else. That's why I'd choose it on internet Forums any time. Simple is good.


The good thing is: It all is there. Anyone can buy lots of different products and pick the ones which fit their wants best.
Sadly I fear none is perfect.

[Big Dave]

We all know if you can eliminate most variables, you can come to a more certain conclusion. Using my A7R which has FF 36mp and asp-c mode 16mp capability. I shot an evenly lite wall in both FF and asp-c, using a 50mm @f2 and a 35mm @f2. All combinations of lenses and formats gave the same ISO, shutter speed @ f2. Since everything is the same, except the format, this should be a fair test. If pixel size and tech level are equal, format size is not a factor. We are talking real world here. Of course a bigger finder does have a brighter image. So I think that better ISO performance is a result of larger pixels.

[osv]

That is a little too little to discuss. What significant "more" is there to be had on a 36x24 crop sensor compared to a 24x15 crop sensor?
Just look at the poor guys who are stuck with a Nikon FF DSLR. They can't achieve what the NEX APSC owner with a Canon 85/1,2 on a wide angle converter can. "More DoF control".

that's apples and oranges, and it's not any kind of a logical comparison... no one who is seriously into p.q. would use a wide angle converter.

The Canon FF DSLR owners don't get the dynamic range the NEX owner gets.
None get the resolving power a NEX user with EF50/1,2 and WAC gets.

that is a boatload of rubbish, lol

good grief.

---------- Post added 11-02-2014 at 08:49 AM ----------

So I think that better ISO performance is a result of larger pixels.

i don't understand why people attempt to argue otherwise, but they do.

"Pixel size. Simply put, the larger the pixel, the more photons reach it, and hence the better the signal-to-noise ratio (SNR) for a given exposure. The number of electrons generated by the photons is proportional to the sensor area (as well as the quantum efficiency). Noise power is also proportional to the sensor area, but noise voltage is proportional to the square root of power and hence area. If you double the linear dimensions of a pixel, you double the SNR. The electron capacity of a pixel is also proportional to its area. This directly affects dynamic range."
Noise in photographic images | imatest

[jsherman999]

i don't understand why people attempt to argue otherwise, but they do.

"Pixel size. Simply put, the larger the pixel, the more photons reach it, and hence the better the signal-to-noise ratio (SNR) for a given exposure. The number of electrons generated by the photons is proportional to the sensor area (as well as the quantum efficiency). Noise power is also proportional to the sensor area, but noise voltage is proportional to the square root of power and hence area. If you double the linear dimensions of a pixel, you double the SNR. The electron capacity of a pixel is also proportional to its area. This directly affects dynamic range."
Noise in photographic images | imatest

The advantage of larger sensors is *not* because the pixel size is larger, but because the sensor area is larger, and thus it gathers more total light for the same exposure.

The portion of the Imatest link you posted is concerned with per-pixel noise. Larger pixels can have better SNR, yes, but the noise in an image isn't determined only by that. We have multiple examples of larger sensors with same-sized or smaller pixels out-performing larger-pixeled, smaller sensors (in image noise comparison.)

[osv]

The advantage of larger sensors is *not* because the pixel size is larger, but because the sensor area is larger, and thus it gathers more total light for the same exposure.

The portion of the Imatest link you posted is concerned with per-pixel noise. Larger pixels can have better SNR, yes,

exactly: " the larger the pixel, the more photons reach it, and hence the better the signal-to-noise ratio (SNR) for a given exposure"

it can't be argued... here it is again:

"“We’re obviously working with very large-frame sensors, 36x24mm sensors that we use in our DSLRs,” said Canon national marketing executive Larry Thorpe. “And with every generation that we come out with, and they come out with some frequency with that domain of the DSLR, we’re always working on noise and dynamic range. Those are the two challenges, especially when you increase the pixel count, because once your photo-site goes down in size, you’re confronted with the challenges of noise and dynamic range.” - See more at: TVTechnology: What?s New in Large Format Sensor Cameras

We have multiple examples of larger sensors with same-sized or smaller pixels out-performing larger-pixeled, smaller sensors (in image noise comparison.)

the only way that you would have an accurate measurement of the pixel size is if you got it off the chip manufacturers data sheet... afaik, pixel pitch is a calculated estimation, but it doesn't necessarily represent pixel size.

i'd like to see actual pixel size, if you have it.

[jbondo]

There are several articles on clarkvision.com that I believe are very relevant to this discussion, and they are well said, and backed up with the maths and science principles.
These two are good starting points.
www.Clarkvision.com: The Depth of Field Myth and Digital Cameras
f-ratio myth

Hopefully this will be a help for some folks to help clear up the numerous misconceptions, including the ones about pixel size scaling with sensitivity, etc. Sometimes the internet becomes a big ol' echosphere of opion and I find sites like this refreshing in that it favors actual science over opinion. Check it out.

[jsherman999]

exactly: " the larger the pixel, the more photons reach it, and hence the better the signal-to-noise ratio (SNR) for a given exposure"

it can't be argued... here it is again:

"“We’re obviously working with very large-frame sensors, 36x24mm sensors that we use in our DSLRs,” said Canon national marketing executive Larry Thorpe. “And with every generation that we come out with, and they come out with some frequency with that domain of the DSLR, we’re always working on noise and dynamic range. Those are the two challenges, especially when you increase the pixel count, because once your photo-site goes down in size, you’re confronted with the challenges of noise and dynamic range.” - See more at: TVTechnology: What?s New in Large Format Sensor Cameras

All they're saying in that link is that they can't increase the # of pixels (decrease the pixel size) indiscriminately without incurring things like read noise. Same reason Sony wouldn't want to come out with an Exmor at 100MP right now. BTW there's no math or physics in that link, just some twice-removed marketing speak

the only way that you would have an accurate measurement of the pixel size is if you got it off the chip manufacturers data sheet... afaik, pixel pitch is a calculated estimation, but it doesn't necessarily represent pixel size.

i'd like to see actual pixel size, if you have it.

?? The pixel size of almost every sensor is well known, available from multiple sources and derivable if you know the sensor size and # of pixels, and places like chipworks measure it directly with microscopy.

For example, D7000 (and K5) and D800 have almost exactly the same pixel size, about 4.8 microns. Now witness:



The D700 is much older and has lower QE than the K5 or D800, yet it still holds an advantage over the smaller sensor - not because it has 'larger pixels' and certainly not 'better pixels', but because the DxO comparison is taken at the same exposure (and thus the FF image is created with more total light.) The D800, of course, has the same pixel size as the K5 and you can see it's Total Light advantage (and the advancement Sony/Nikon have made since the D700 came out.)

You can see some advantages to bigger pixels if you find an example of a manufacturer trying to maximize SNR with the current tech available, like for example the D4s, but the difference between that and for example the D800 isn't huge (much smaller delta than D800 vs. any aps-c sensor) and it's not attributable to the consistent advantage you see in larger vs. smaller sensors - the big difference between aps-c and FF of the same generation for example is due to Total Light, not size of the pixels.

(It's also important to realize the difference between per-pixel performance and overall image, and when something being cited is talking about one vs. the other.)


.

[jsherman999]

There are several articles on clarkvision.com that I believe are very relevant to this discussion, and they are well said, and backed up with the maths and science principles.
These two are good starting points.
www.Clarkvision.com: The Depth of Field Myth and Digital Cameras
f-ratio myth

Hopefully this will be a help for some folks to help clear up the numerous misconceptions, including the ones about pixel size scaling with sensitivity, etc. Sometimes the internet becomes a big ol' echosphere of opion and I find sites like this refreshing in that it favors actual science over opinion. Check it out.

I think that's a good source too for some things, but know that some folks have some problems with some of his verbage on pixel size. (I think he's amended it some from the early days, perhaps, taking some new thinking into account?) Anyway, take this as a caveat to consider: The problem with Dr Clark's narrative.

.