Full frame vs. cropped sensors
- Thread starter Amfibius
- Start date
- Status
Watcher said:
By equal measure, you cannot be comparing cameras separated by 2 years and a generation of technology. It's like saying my 2003 1.6L Subaru has more power than my Dad's 2.0L 1990 Volvo, therefore there is no correlation between engine capacity and engine power. Patently false! Compare like for like, take a 2002 2.0L Volvo S60 and compare with the 2003 3.0L Merc E class, same generation of technology, similar implementation, then we're talking.
dkw, what I was expecting from amfibius is just a simple reasoning to relate pixel pitch to DR. When simply stated it just requires 1 answer, greater full well capacity assuming construction materials are similar, which nicky provided as an analogy earlier and you have provided now as well.
I hope you are aware that what I'm after is reasoned points and not dumping info.
I hope you are aware that what I'm after is reasoned points and not dumping info.
Zerstorer, Watcher, just forget it. You have lost. Larger pixels = postential for increased dynamic range. FACT.
How many more links do you want us to post? Can't either of you see that you are painting yourselves into a corner? Your position is wrong, just admit it now and forget it - let bygones be bygones. If you keep on posting, you are defending an untenable position and you only undermine your own credibility.
As for simple reasoning of how pixel pitch relates to dynamic range, I did do that - see the water bucket analogy. How much simpler do you want me to get? Like I said, I can't help it if you still can't see my point. That is about as simple as I know how to get without having to teach you that A is for Apple, B is for Bee, C is for Cat ...
How many more links do you want us to post? Can't either of you see that you are painting yourselves into a corner? Your position is wrong, just admit it now and forget it - let bygones be bygones. If you keep on posting, you are defending an untenable position and you only undermine your own credibility.
As for simple reasoning of how pixel pitch relates to dynamic range, I did do that - see the water bucket analogy. How much simpler do you want me to get? Like I said, I can't help it if you still can't see my point. That is about as simple as I know how to get without having to teach you that A is for Apple, B is for Bee, C is for Cat ...
Zerstorer said:
*Snip* [Resorting to this and refusing to reply simply shows that you were simply regurgitating hearsay and false claims without any understanding to the subject matter and being unable to explain the logic behind your claims.
Rather juvenile response don't you think?]
Nothing wrong with quoting information if it is established fact, so as to move the discussion along. If he had mentioned that larger photo-sites have bigger DR, it is based on fact, it is not hearsay nor false claims.
If I were trying to teach nuclear physics to my son, and if I say the nucleus of an atom comprises protons and neutrons, and this is based on fact, then it does not make my claim any less valid if I cannot explain WHY the nucleus is so. I don't have to be a nuclear physicist to be able to teach some basic physics. Likewise, his claim about larger photo-sites is valid, as you have now conceded. If he had not explained WHY, it doesn't really matter, because it is established and common fact and your challenge on the basis of "reasoned points" is a false one. If we have to derive everything from first principles every time we have a discussion, how on earth do we progress?
Ok, if you don't want 2 year separation of technology, 1D MkII has a pixel size of 8.8micrometer, 1Ds MkII has 7.2 micrometer. Both Canon sensors are using the same technology and are the same generation and are pro system.dkw said:
Now since 1D MkII has the approximate the same dynamic range and color range as the 1Ds MkII but is 8.2/7.2 about 14% bigger than the 1Ds MkII, then is there a relationship there?*
What is wrong with my calculation, dkw and my dear Amfibius. Both sensors are the same generation, come from the same manufacturer, separated by oh 3-6 months only. Why, by your view, the 1DsMkII should have significantly lesser dynamic range and color range, right?
* from http://www.imaging-resource.com/NEWS/1095750074.html
Noise, we agree, but dynamic range and color depth? At around 86% of the size for 1Ds Mk II, there would be noticable reduction.Amfibius said:
If you take the older models, and compare the digital back with a DSLR, the issue holds out consistently. The H20 has a pixel size of 9 micrometer as well. Now, the 1Ds which is has 8.8 micrometer which is newer by 8-9 months (both are 2002 products). H20 again has 16-bit color depth, 12 f-stop on its dynamic range vs 12 bit color, 6-7 f-stops dynamic range. 2% bigger for the H20 but so much better in terms of performance. Why?
All that techno mumble-jumble is confusing me.
I didn't read it all, but I guess I more of less got the gist of it.
I think we all agree that FF sensor would be a nice thing to have, better to capture all that is possible, than a cropped image sensor that lets light and image fall onto nothingness.
Comparing APS-sized sensors to 35mm-sized sensors is a little less removed than comparing 35mm-sized sensors to medium format backs.
A BMW that is costly to a Corolla owner is loose change to a Ferrari owner.
Don't diss a BMW just because you can't afford it(or you can afford much more than it), assess it objectively.
I didn't read it all, but I guess I more of less got the gist of it.
I think we all agree that FF sensor would be a nice thing to have, better to capture all that is possible, than a cropped image sensor that lets light and image fall onto nothingness.
Comparing APS-sized sensors to 35mm-sized sensors is a little less removed than comparing 35mm-sized sensors to medium format backs.
A BMW that is costly to a Corolla owner is loose change to a Ferrari owner.
Don't diss a BMW just because you can't afford it(or you can afford much more than it), assess it objectively.
Watcher said:
Guffaw***,
By your logic, the Sony 8MP sensor on the F828 should have the same DR as the 20D......
Its not only MY view, did you read any of the links? Everybody else is wrong except you? Go argue with DPReview and British Journal of Photography. Better yet, post this on DPR, Fred Miranda and Rob Galbraith, "Watcher says, photo-diode size has no practical impact on dynamic range, therefore, FF is not necessary." Go ahead, do it.....PROVE your point.
This is called "putting words into people's mouth" and suffering from logic failure. You seem to suffer from one of the Bushism: You are either with us or against us.dkw said:
A =/=> B does not mean !A ==> !B.
Please show me where I said FF is not necessary (for all situation). Please. I have shown real products that can be bought that shows that pixel size even from the same manufactuer within a period of 6 months, does not have a correlation to the dynamic range and color range. What do you do?
True, but on the other hand, don't look down on Corolla even if you can afford a BMW. That is call humility.r32 said:
More comments to this post.dkw said:
I had met your challenge in your post #82 where you said:
dkw said:
I matched digital sensors from the same manufacturer (Canon) and in the same class (Pro) and within a couple of months. The difference is 86% in size. They have the same color range (12-bits) and dynamic range.
Even the H20 while has only 2.2% larger sized pixels, has dramatically larger dynamic range and color range when compared to the 1Ds which was released more than 6 months after the H20 comes out.
So now what? You seem to resort to insults and empty challenges and you move the goalpost after someone has suceeded in matching your challenge.
i wud like to put this as objective as possible. this is really just a case of theory vs reality from what i see.
in theory, a bigger pixel pitch yields bigger DR. however in reality, we forgot that by making/tweaking the properties of the pixel, we can achieve higher efficiency from it.
i do believe there's a compromise curve that sensor engineers know. by having modern tweaking & manufacturing process, making the pitch larger probably didn't have too significant impact on the noise & DR. so if u can have something smaller, better, last longer & cheaper, why go for the primitive way of doing it or try to be close to those theories? let me put in a few analogies to emphasis my point:
1) transistors. the transistors of the 50s & 80s. both semi conductors but as manufacturing processes improve, products improve in terms of performance, reliability & cost. does it mean bigger is better?
2) filament light bulbs in the 1900s. i think today, your mag light can outlast it & outshine it. both r using filament bulbs. bigger is better? of cos u can argue that they r centuries apart but thats not what i want to illustrate here.
3) engine capacity. if the 1.5kcc car is built smaller & lighter with higher engine efficiency, i don't think it will be anywhere inferior to a 2.0kcc car. my friend told me that a 1.0kcc car can't go up Genting cos of the steep slope, but hey, look at those 1.0kcc motorbike, very powerful.
4) big bucket vs small bucket. yes, u can have big buckets to collect more water, thats the theory anyway. but if u use a small bucket & put a big funnel on it, u can collect as much water as a big bucket. not only can u collect more water with a funnel, but by making the funnel from silicon dioxide material, u can make it resistant to water sticking on it (cohesivity?). u can further refine it by coating a layer of teflon on it to make it "more accurate" with directing water flow.
coming back to the topic, the no. of pixel is limited by the size of the sensor, so there's a limited no. of big pixel u can sqeeze. but by putting a micro lens on a smaller pixel, u improve its efficiency & can gather the same amount of light as a big pixel w/o micro lens. further improvement in the pixel material & micro lens can push u even further. so by now, does everyone get what i m saying? costly or not, money is always an issue, so lets not compare who in this club has deeper pockets.
looking at the transistor analogy again, the theory behind transistor is true but does it mean that the big, clunky transistor made by mr. schottky is better than those found in a P4 (not talking about vacuum tubes vs semi-con)? so its really more of a theory vs reality issue. i hope nobody gets upset becos of this thread.
:cheers: its tiger time!!
in theory, a bigger pixel pitch yields bigger DR. however in reality, we forgot that by making/tweaking the properties of the pixel, we can achieve higher efficiency from it.
i do believe there's a compromise curve that sensor engineers know. by having modern tweaking & manufacturing process, making the pitch larger probably didn't have too significant impact on the noise & DR. so if u can have something smaller, better, last longer & cheaper, why go for the primitive way of doing it or try to be close to those theories? let me put in a few analogies to emphasis my point:
1) transistors. the transistors of the 50s & 80s. both semi conductors but as manufacturing processes improve, products improve in terms of performance, reliability & cost. does it mean bigger is better?
2) filament light bulbs in the 1900s. i think today, your mag light can outlast it & outshine it. both r using filament bulbs. bigger is better? of cos u can argue that they r centuries apart but thats not what i want to illustrate here.
3) engine capacity. if the 1.5kcc car is built smaller & lighter with higher engine efficiency, i don't think it will be anywhere inferior to a 2.0kcc car. my friend told me that a 1.0kcc car can't go up Genting cos of the steep slope, but hey, look at those 1.0kcc motorbike, very powerful.
4) big bucket vs small bucket. yes, u can have big buckets to collect more water, thats the theory anyway. but if u use a small bucket & put a big funnel on it, u can collect as much water as a big bucket. not only can u collect more water with a funnel, but by making the funnel from silicon dioxide material, u can make it resistant to water sticking on it (cohesivity?). u can further refine it by coating a layer of teflon on it to make it "more accurate" with directing water flow.
coming back to the topic, the no. of pixel is limited by the size of the sensor, so there's a limited no. of big pixel u can sqeeze. but by putting a micro lens on a smaller pixel, u improve its efficiency & can gather the same amount of light as a big pixel w/o micro lens. further improvement in the pixel material & micro lens can push u even further. so by now, does everyone get what i m saying? costly or not, money is always an issue, so lets not compare who in this club has deeper pockets.
looking at the transistor analogy again, the theory behind transistor is true but does it mean that the big, clunky transistor made by mr. schottky is better than those found in a P4 (not talking about vacuum tubes vs semi-con)? so its really more of a theory vs reality issue. i hope nobody gets upset becos of this thread.
:cheers: its tiger time!!
Watcher said:
Watcher,
I actually have a hard time trying to understand what it is exactly you are trying to say, so I paraphrased. I ask only one thing, read the links, and explain to me why photo-diode size has no relationship to DR.
Your comparisons between 1Dmk2 and 1DsMk2 make too many assumptions and are seriously flawed.
Assumption no.1: 1dsMk2 has more DR than 1dMk2. Well, this a proven fact, or based on 1 person's review?
Assumption no.2: 1dsMk2 sensor is "only" 3-6 months newer than 1dMk2. Unless you belong to the Canon R&D arm, I'm not sure how you would know that. For all you know it could be an 18 month older sensor.
Assumption no.3: FF vs 1.3x crop sensor will have same "electronic" characteristics. Admittedly, I'm also making an assumption here. It could be that for a larger sensor there is more leeway to design the circuitry such that the output from each photo-diode can be drained away more quickly.
The comparison you have put up is basically too shaky to have any credence. Yes, the 1ds pixel is 14% smaller, to you, that's a significant decrement, to me, it's not. There are simply too many variables. Take for example the 10D and 300D sensor, to all intents and purposes the same sensor. Did you see the DPR noise charts for those, 10D better than 300D! How could that be? Well, utterly possible! Tolerances in production, batch differences all contribute to variability. Again I will use an automobile example, I used to drive a Ford Laser, a twin of the Mazda 323. However, the 1.6L engine of the Laser, which came from the same factory as the Mazda, was rated at 3bhp higher! Can you then use that to draw the conclusion that engine capacity has no relation to bhp? Obviously not! Yet, that's what you keep doing over and over and over. Variability matters, and the example you cite, simply too many unknowns to support your contention. In fact, if we were to use your methodology, the only logical conclusion that can be derived, since you have proven beyond all doubt that a smaller photosite has SIMILAR DR as a larger one (this is what you are saying, yes?), is that all camera makers should quit FF and APS, and only make P&S sized sensors.
No Bushisms, I don't care for FF, Nikon or Canon. Don't see that I will ever buy FF and if Canon fall behind, I'll be the first to change.
In a nutshell, you are trying to say that photo-diode size has no impact on DR, and support your argument by quoting a few examples of released cameras with smaller photo-diodes which have better DR than bigger photo-diodes. If I were to use your methodology, for every example you quote, I can quote another 5 examples showing the opposite (DR: 10D > A80, 300D > G5, D70 > CP 8700....get the picture), then who's right? Your methodology is seriously flawed, doesn't prove anything and doesn't get to the guts of the issue.
The point I am trying to emphasise, which you fail to see, and I haven't moved any goal posts, by the way, is that for the same conditions, and larger photo-diode will have more DR than a smaller one. That's all.
My last post on this. The challenge on posting elsewhere is not an empty one, I will be happy to link this to the other forums if you are agreeable, it will be good to hear some knowledgeable opinions.
Last, I don't see how I have insulted you, but I do apologise if you feel that you have been insulted, not at all my intention. As you see from my posts here I only entered the guts of this discussion late, when some quite unreasonable things were being said to Amfibius.
Cheers,
Well, I've sure learned a thing or two from this entire argume...errr, discussion, though at times it seemed like it was leading into another brand war, the gist of the argument seems to have prevailed.
In the interest of mutual learning, I have linked this thread to DPReview's Pro forum:
DPReview ProDigital Forum
Cheers
GYR
In the interest of mutual learning, I have linked this thread to DPReview's Pro forum:
DPReview ProDigital Forum
Cheers
GYR
First let me apologise for starting this and then disappearing, but I was unavoidably away from the Internet for several days longer than I had anticipated.
The discussion has, for this kind of sensitive topic, stayed *fairly* polite and on track, although it has largely degenerated into a full frame sensors have more potential for more dynamic range therefore they are better than 1.5x crop sensors. Which is not true. They might have better dynamic ranger, but there might be other reasons why a 1.5x crop sensor is better, which may or may not make it better overall. So don't get caught in the full frame = better dynamic range = better generally whirlwind.
In response to the dynamic range issues, I agree with the pro FF front; in *theory* more sensitive photosites equates to less noise, which translates to potentially better dynamic range as manufacturers can dig further into the shadow regions to produce detail without detrimental effect to noise generated.
However, as has been conveniently ignored, I did point out in my first post that similarly in *theory*, lenses made for a smaller sensor can be similarly faster, resulting in increased light transmission, and cancelling out any gain in sensitivity on the part of the sensor. Put another way, the average point and shoot digital with respectable tele lens has a very respectable maximum aperture, usually in the f2.8-f4 range. Compare this with say f13 on the Pentax Espio that once upon a time boasted the longest telephoto zoom lens on a point and shoot film camera. It's not a perfect analogy, but I'm just trying to illustrate what I mean. The gains won't be as big, but it's the theory behind it I'm aiming at.
Once you accept this position (it is only theory, because the lenses haven't appeared, but then the signal response argument put forward by the pro-FF people is also largely theoretical), then the FF sensor having more dynamic range point loses its authority.
Regarding format sizes, I wouldn't consider 35mm the "sweet spot". As has been pointed out, at one stage in history, 10x8 was the sweet spot, and people cringed at 5x4. Same happened with 6cm roll film arrived. And the same happened when 35mm replaced medium format as the main format. Resistance to change and attachment legacies will always cause people to react to that change.
Also, it has been claimed in this thread that FF has more DOF control than a 1.5x sensor, ceteris paribus. That statement simply isn't true. It is really a double edged sword. With the bigger sensor you gain the ability to narrow and restrict the depth of field. But you also lose the ability to broaden the depth of field at the other end of the scale. I've heard so many people go on about how they cannot get as out of focus backgrounds with the DX sensors, compared to their full frame shots. But everyone also conveniently forgets that you also struggle to get enough DOF in your full frame shots, compared to what you get with a 1.5x sensor. Given a telephoto shot that displays slightly better out of focus backgrounds, versus a landscape shot with not all its important elements in focus, I know which shot would be a better compromise.
Also, back to the lens design argument I'm afraid, but if you again accept the (theoretical) ability to design faster lenses with a smaller sensor, the ability to isolate DOF becomes improved.
As to FF placing less resolution demand on glass than a smaller sensor, that is true. But again as I've pointed out in my original post, it places greater demands on correcting distortion and aberrations particularly at the edges. Which means that it then becomes possible to design better lenses for the smaller sensor, resolution wise. This is what Olympus and Nikon have been claiming for a long time.
Streetshooter, regarding sensor arrays, the day already arrived several years back with the EOS 1D, which is really two 2mp sensors with separate data readouts and merged together at the processing stage. Not quite 4, 9 and 16, but along the lines...
As to the point that most people against full frame seem to be Nikon users, first let me state again that I am NOT against full frame. I am against the perception that the DX sensor is something to be ditched at the earliest available opportunity. I am at least as happy with DX as I am with full frame, but there are plenty out there who also think that DX is a serious handicap; I disagree with that notion.
Secondly, can I also point out that [1] people who point out that people against full frame are Nikon users are Canon users and [2] people who are pro FF also tend more often than not to be Canon users. Doesn't mean that "something's really wrong".
As has already been explained by 2100, it is very important to distinguish between bit depth and dynamic range. And also in all this discussion that the Nikon D1x, ancient camera that it is, was the dynamic range "king" in its time, and it captured about 3.2D worth of information. And then you need to realise that paper doesn't display that range to begin with, so the moment you print a *properly exposed* digital camera file, you are probably throwing away (or more accurately compressing) your available dynamic range. Ditto 16 bit files; printers only print in 8 bit, so while maintaining 16 bit files allow greater flexibility when editing, assuming you capture an 8 bit file properly, the resultant print is still going to be very good. Assuming you consider current digital printing technology adequate. The point being, yes a full frame sensor offers the potential of better dynamic range with the same amount of light reaching the sensor; but we have also got a bigger bottleneck elsewhere in the workflow.
(Cont'd)
The discussion has, for this kind of sensitive topic, stayed *fairly* polite and on track, although it has largely degenerated into a full frame sensors have more potential for more dynamic range therefore they are better than 1.5x crop sensors. Which is not true. They might have better dynamic ranger, but there might be other reasons why a 1.5x crop sensor is better, which may or may not make it better overall. So don't get caught in the full frame = better dynamic range = better generally whirlwind.
In response to the dynamic range issues, I agree with the pro FF front; in *theory* more sensitive photosites equates to less noise, which translates to potentially better dynamic range as manufacturers can dig further into the shadow regions to produce detail without detrimental effect to noise generated.
However, as has been conveniently ignored, I did point out in my first post that similarly in *theory*, lenses made for a smaller sensor can be similarly faster, resulting in increased light transmission, and cancelling out any gain in sensitivity on the part of the sensor. Put another way, the average point and shoot digital with respectable tele lens has a very respectable maximum aperture, usually in the f2.8-f4 range. Compare this with say f13 on the Pentax Espio that once upon a time boasted the longest telephoto zoom lens on a point and shoot film camera. It's not a perfect analogy, but I'm just trying to illustrate what I mean. The gains won't be as big, but it's the theory behind it I'm aiming at.
Once you accept this position (it is only theory, because the lenses haven't appeared, but then the signal response argument put forward by the pro-FF people is also largely theoretical), then the FF sensor having more dynamic range point loses its authority.
Regarding format sizes, I wouldn't consider 35mm the "sweet spot". As has been pointed out, at one stage in history, 10x8 was the sweet spot, and people cringed at 5x4. Same happened with 6cm roll film arrived. And the same happened when 35mm replaced medium format as the main format. Resistance to change and attachment legacies will always cause people to react to that change.
Also, it has been claimed in this thread that FF has more DOF control than a 1.5x sensor, ceteris paribus. That statement simply isn't true. It is really a double edged sword. With the bigger sensor you gain the ability to narrow and restrict the depth of field. But you also lose the ability to broaden the depth of field at the other end of the scale. I've heard so many people go on about how they cannot get as out of focus backgrounds with the DX sensors, compared to their full frame shots. But everyone also conveniently forgets that you also struggle to get enough DOF in your full frame shots, compared to what you get with a 1.5x sensor. Given a telephoto shot that displays slightly better out of focus backgrounds, versus a landscape shot with not all its important elements in focus, I know which shot would be a better compromise.
Also, back to the lens design argument I'm afraid, but if you again accept the (theoretical) ability to design faster lenses with a smaller sensor, the ability to isolate DOF becomes improved.
As to FF placing less resolution demand on glass than a smaller sensor, that is true. But again as I've pointed out in my original post, it places greater demands on correcting distortion and aberrations particularly at the edges. Which means that it then becomes possible to design better lenses for the smaller sensor, resolution wise. This is what Olympus and Nikon have been claiming for a long time.
Streetshooter, regarding sensor arrays, the day already arrived several years back with the EOS 1D, which is really two 2mp sensors with separate data readouts and merged together at the processing stage. Not quite 4, 9 and 16, but along the lines...
As to the point that most people against full frame seem to be Nikon users, first let me state again that I am NOT against full frame. I am against the perception that the DX sensor is something to be ditched at the earliest available opportunity. I am at least as happy with DX as I am with full frame, but there are plenty out there who also think that DX is a serious handicap; I disagree with that notion.
Secondly, can I also point out that [1] people who point out that people against full frame are Nikon users are Canon users and [2] people who are pro FF also tend more often than not to be Canon users. Doesn't mean that "something's really wrong".
As has already been explained by 2100, it is very important to distinguish between bit depth and dynamic range. And also in all this discussion that the Nikon D1x, ancient camera that it is, was the dynamic range "king" in its time, and it captured about 3.2D worth of information. And then you need to realise that paper doesn't display that range to begin with, so the moment you print a *properly exposed* digital camera file, you are probably throwing away (or more accurately compressing) your available dynamic range. Ditto 16 bit files; printers only print in 8 bit, so while maintaining 16 bit files allow greater flexibility when editing, assuming you capture an 8 bit file properly, the resultant print is still going to be very good. Assuming you consider current digital printing technology adequate. The point being, yes a full frame sensor offers the potential of better dynamic range with the same amount of light reaching the sensor; but we have also got a bigger bottleneck elsewhere in the workflow.
(Cont'd)
- Status
Similar threads
