Full frame vs. cropped sensors

[reflecx]

Nothing to buying a 300/4, and mount it on your FF camera, and then cropping off half the picture and faking a "600/4" lens is there? In fact, you still get higher resolution if you crop a 1DsMk2 when compared to the 2x crop of the D2X: 16MP/2 = 8MP on the 1DsMk2, vs. 6.8MP on the D2X (if I remember correctly). So unfortunately I cannot agree with you on this point.

* cough cough * sorry to be cynical, but if this shows the extend of your knowledge ....

You don't divide by the the crop factor, you divide by the crop factor squared, or 16.7/4 = 4.2.

 

[Zerstorer]

Listen, I can't be bothered with you any more. If you want me to demonstrate proof, of course I can. But you need to make it worthwhile for me. If I can demonstrate proof, will you run naked around the Merlion and have pictures posted up on Clubsnap?

Really at this again? Shows much of your character.

 

[dkw]

You choose not to answer but to insult. Do you really know what you are claiming?


This is totally irrelevant to your claim. If you can pull this out I can see how desperate you are to avoid the topic.

The final image depends on the processing circuitry and algorithms other than the output from the sensor. You have to prove beyond doubt that the sensor captures greater dynamic range in the first place.

You are the one making a claim. Show conclusive proof(not analogies nor irrelevant examples) of it, or otherwise acknowledge that its just an assumption.

Zerstorer, you are a pseudo-scientist.....of course he can't prove it beyond all doubt, he's not an sensor engineer and neither are you, but there is more than enough circumstantial evidence that a larger photosite will have more dynamic range than a smaller one, INDEPENDENT of whatever circuitry or algorithm you may put in place.

Read this article from the British Journal of Photography---> http://db.riskwaters.com/public/showPage.html?page=13880

The fifth paragraph reads "...is a solution to the larger dynamic range of larger photo-diode site...... A bigger single photo-diode site can offer greater sensitivity and dynamic range than a a smaller one..."

Game, set and match to Amfibius. How about an apology Zerstorer?

 

[Watcher]

Price should not be the issue here.

Before we go on, i think we should make it clear that we can approach this FF-APS thing from two fronts: the physics/engineering aspect, and the practical/ecomonic aspect.

But it is and can not be totally removed from any debate. If is not the issue, everyone will buy Hummers (vehicles, not birds ) already. But I do agree that we should look from two different angle.

From the engineering perspective, we need to strip out the following factors:

- Brand. It's just five letters on the camera body right? Nikon can FF too; they chose not to, up to now.

I have gone on record to say that if it makes economic sense and if I have the need, I will get a FF DSLR. Precisely agree on that Nikon can make FF. What makes people (esp opposing brand supports who seems to suffer from projection issues) think that Nikon can't? There is a vast difference between won't and can't.

- Pixel size. See my previous post.

As said by Zerstorer, pixel size does not imply better quality. 1DMkII and 20D has significantly smaller pixel size. Are they significantly worse in terms of quality vs 1D and D60? There are more issue at hand than that.

- Price. At least from a technical aspect, this is not relevant. FF are more expensive, but they are not in the $120,000 range (eg 1200/5.6L). Put in perspective, a car in Singapore can lose $12k in about 3 years or so. Just 5 years back, before the D30 came along, dSLRs were all way above S$12k. The price of FF will come down. If you want to say that it costs more than APS cameras, i concede the point. But it is not ridiculously out of reach. Price/performance ratio is a subjective and personal thing, so let's not go there.

Though it is not a technical issue. It is relevant to the discussion. Why? Do you make purchase decision independant of price? From what I know, even Bill Gates calculates how he spends his money. For amateurs (defined as those who do not draw significant income from) photographers, it makes it even more important. Similarly, pros (or their company) do calculate ROI. Out of reach? No, but does it make sense after you compare the advantages vs price? That is my point.

<chopped>

Approaching from the economic-sense perspective, things get muddier. A FF (of either brand, or any brand) is nice to have, as many would attest, but not everyone is willing to dole out the dough. i don't we'd ever reach a concensus on this and i have no further comments on this part.

Personally i feel that buying a 300/2.8 so that i can get an effective 450/2.8 or a 600/2.8 doesn't make sense, from a FF vs APS viewpoint. Why not get a fullframe and crop it to the desired picture? i paid good money for the 36x24 image. Especially the outer edges. It makes sense when your APS sensor has DENSER pixels than the fullframe one, but when the density is eventually the same, you'd just be throwing a huge part of a good image away.

But if by using a reduced size sensor, the performance of the camera is faster and needs less initial storage space say on the CF card, is it not an advantage that makes sense?

Streetshooter, i'm sure someone thought of the 'joining sensors' idea. i believe in the interview with the 1Ds design team, they mentioned that the spacing between pixels was subjected to some incredibly small tolerance, beyond which they had to reject the sensor. Maybe someone *WILL* find a way to stitch sensors; when that day comes, we'd all have cheap FF sensors and be laughing our socks off at silly 'FF vs APS' discussion threads.

Old, old news. Someone tried and failed on this 7-8 years ago. Hint: Minolta and RD-175 (I think).

 

[dkw]

More stuff,

http://www.machinevisiononline.org/public/articles/Mark_Butler.pdf

Slide 16 reads....

Robots and Vision 2003
CMOS Imaging and Application in Machine Vision
Impact of Shrinking Design Rules
Forced upon machine vision by foundries
satisfying the needs of the consumer market
• leads to greater pixel resolutions in
smaller packages
• leads to smaller pixel sizes (3um-5um)
• leads to lower operating voltages (2.5V or
3.3V)
• leads to poorer dynamic range

Slide 19 reads....

Robots and Vision 2003
CMOS Imaging and Application in Machine Vision
Impact to Dynamic Range
• Lower operating voltages reduce the
output voltage of the sensor amplifier
hence the electron storage capacity (full
well) of the device becomes smaller
• A reduction in full well leads to poorer
signal to noise or dynamic range because
shot noise reduces at a slower rate
• The relationship of full well to lower
operating voltages is sub-linear

Slide 21....

Robots and Vision 2003
CMOS Imaging and Application in Machine Vision
Conclusions
• There is a trade off between speed and
windowing (parallel, single sloped ADC vs.
serial ADC)
• ~7um pixel size provides good MTF, consider
this when selecting smaller pixel devices
• Shrinking design rules improve fill factor, but
lower operating voltages decrease dynamic
range
• Consider S:N ratios when using small full wells



From a marketing spiel...

http://www.pulnix.com/Imaging/i-techA.html

Quote--> "The design of the 1" imager used in the PULNiX cameras will not saturate as quickly as the 2/3" imager used in the other cameras, allowing a higher dynamic range."

Zerstorer?

 

[Amfibius]

* cough cough * sorry to be cynical, but if this shows the extend of your knowledge ....

You don't divide by the the crop factor, you divide by the crop factor squared, or 16.7/4 = 4.2.

Thanks for the correction.

 

[espn]

Listen, I can't be bothered with you any more. If you want me to demonstrate proof, of course I can. But you need to make it worthwhile for me. If I can demonstrate proof, will you run naked around the Merlion and have pictures posted up on Clubsnap?

If you don't demonstrate proof, how to let people trust & believe in you?

It's like asking a whole legion of soldiers to march into the raging sea waters when you (assuming the role of an officer) tells them to just do it and you stand way behind them tell them/assuring them its safe to do so as long as you can swim?

And without proof how to substantiate your claims? This is meant to be an open discussion and you *are* taking it a bit too personal. How to have a discussion when one doesn't want to proof their claims/findings and expects the 'challenger' to strip and run around naked if proven wrong?

Do accept things gracefully and peacefully, if you can't take any questions and 'challenges' then don't start this thread in the first place.

 

[Amfibius]

Game, set and match to Amfibius. How about an apology Zerstorer?

Thanks for your help in digging up those articles dkw. Not that I want to undermine your efforts, but I managed to find them just as easily with Google. I had all those links bookmarked at home, just waiting for Zerstorer to accept my challenge to run naked around the Merlion after I post them

I think what I wanted to emphasize was that it's easy enough to do the research yourself on Google rather than get on this forum and demand until you are red in the face for something as easily demonstrable as that. This is not good debate tactics and it would have been very easy to skewer him alive if you waited a bit to let him dig himself a deeper hole before furnishing the proof.

Don't need an apology from anyone. I have been as inflammatory as anyone else in this thread. He might feel offended, but I'm not losing any sleep over it.

 

[dkw]

If you don't demonstrate proof, how to let people trust & believe in you?

It's like asking a whole legion of soldiers to march into the raging sea waters when you (assuming the role of an officer) tells them to just do it and you stand way behind them tell them/assuring them its safe to do so as long as you can swim?

And without proof how to substantiate your claims? This is meant to be an open discussion and you *are* taking it a bit too personal. How to have a discussion when one doesn't want to proof their claims/findings and expects the 'challenger' to strip and run around naked if proven wrong?

Do accept things gracefully and peacefully, if you can't take any questions and 'challenges' then don't start this thread in the first place.

I really don't know why I'm allowing myself to get sucked into this, but you guys are just haranguing him. What he stated is conventional wisdom and well accepted POV, but you insist on proof.....he's not a sensor engineer, how does he "prove" it? I can ask you to prove from first principles that the atmosphere comprises 21% oxygen, a conventionally accepted POV, can you? I hope my 2 posts above have put your doubts to rest that a larger photo-site inherently has more dynamic range than a smaller one.

 

[dkw]

Thanks for your help in digging up those articles dkw. Not that I want to undermine your efforts, but I managed to find them just as easily with Google. I had all those links bookmarked at home, just waiting for Zerstorer to accept my challenge to run naked around the Merlion after I post them

I think what I wanted to emphasize was that it's easy enough to do the research yourself on Google rather than get on this forum and demand until you are red in the face for something as easily demonstrable as that. This is not good debate tactics and it would have been very easy to skewer him alive if you waited a bit to let him dig himself a deeper hole before furnishing the proof.

Don't need an apology from anyone. I have been as inflammatory as anyone else in this thread. He might feel offended, but I'm not losing any sleep over it.

Shucks! Could have made it a CS outing at the Merlion :bsmilie: . Okay, get back to work and stop being inflammatory......

 

[Watcher]

Before any x-rated events happen and you guys pat yourself sore in the back, take a look at the pixel size of 1) 1D, 2)1Ds and 3) PhaseOne's pixel size vs color depth and dynamic range. We all do agree that 1D and 1Ds roughly has the same dynamic range and color depth (7-8 f/stops and 12 bit).

That is what Zerstorer is talking about. Yes, pixel size can affect but not a correlation between pixel size and dynamic range and color depth.

The other documents dkw showed are all theoretical issues. Practically, they can be resolved. If PhaseOne can do it, why can't Canon, being such a leader of imaging technology can't ? Or is that won't?

I have shown actual implementation that showed that pixel size not related to color depth and dynamic range. So what now?

 

[Zerstorer]

Zerstorer, you are a pseudo-scientist.....of course he can't prove it beyond all doubt, he's not an sensor engineer and neither are you, but there is more than enough circumstantial evidence that a larger photosite will have more dynamic range than a smaller one, INDEPENDENT of whatever circuitry or algorithm you may put in place.

Note that I have already mentioned that I have no disagreement if it's stated ceteris paribus.

My beef is with him logic, throwing insults, changing his definitions and making very broad claims without substantiation.

What I would term a well reasoned argument is ST1100's points. I have no problem with that as he applies them within appropriate clauses.

Game, set and match to Amfibius. How about an apology Zerstorer?

I was asking him to explain in his words his claims, to support his arguments. One doesn't simply state something without backing it up with reasoning.

I was substantiation all along, yet he didn't provide it at every juncture. How does having you provide it for him help?

Just regurgitating stuff that you trawled up now doesn't prove anything.

 

[Kit]

I think people here have way too much time on their hands which can be efficiently utilised in process of actual image making.

 

[Zerstorer]

I hope my 2 posts above have put your doubts to rest that a larger photo-site inherently has more dynamic range than a smaller one.

You have just made another wild claim.

If the above is true a D1 will undoubtably have more dynamic range than a 1Dmk2.

See my point about quantifying ones arguments?

 

[Amfibius]

Watcher,

My previous post

Let's put it this way ... sticking stubbornly to the 1.5X crop is like a car manufacturer saying that they like 1.5L engines and will produce nothing but cars with 1.5L engines. If they need a car which produces more power - no problem, they'll just make it rev higher, or add a turbocharger or something.

Eventually a limit will be reached when not much further power can be derived from the engine without an increase in the displacement, which will be hard to do because you have convinced all your fanboys that 1.5L engines is the way to go.

I think we both agree that pixel pitch is just one factor which can influence DR, NR, etc. All things being equal, a full frame sensor will deliver cleaner images than a smaller sensor.

 

[Amfibius]

Just regurgitating stuff that you trawled up now doesn't prove anything.

Oh Denial ... is a riiiiiver ... in Eeeeegypt ... :bsmilie: :bsmilie: :bsmilie:

 

[Watcher]

You have just made another wild claim.

If the above is true a D1 will undoubtably have more dynamic range than a 1Dmk2.

See my point about quantifying ones arguments?

:thumbsup: Another example is the 1D has a bigger pixel size than a 1Ds and a P25. Yet, it has a smaller dynamic range and color depth than the P25 and the same as the 1Ds. So how?

Or that the 20D, 1DMkII and 1Ds MkII has smaller pixel size than their predecessors. Are you tell us that they have smaller dynamic range than 10D, 1DMkII and 1Ds Mk II respectively? Then what about all the noise and mutual back patting in here?

 

[Watcher]

Watcher,

My previous post

Let's put it this way ... sticking stubbornly to the 1.5X crop is like a car manufacturer saying that they like 1.5L engines and will produce nothing but cars with 1.5L engines. If they need a car which produces more power - no problem, they'll just make it rev higher, or add a turbocharger or something.

Eventually a limit will be reached when not much further power can be derived from the engine without an increase in the displacement, which will be hard to do because you have convinced all your fanboys that 1.5L engines is the way to go.

For sure, if the 1.5L car can provide all I need and most of my wants at a reasonable price, why not? Unless the 3.0L or 5.0L can satisfy needs/wants that is worth more than the difference in initial price and operation price/cost/issue, why would I want/need to buy one? A 3.0L/5.0L/FF might be superior in some aspect but definately not in all (eg fps, cost of owning one, etc).

I think we both agree that pixel pitch is just one factor which can influence DR, NR, etc. All things being equal, a full frame sensor will deliver cleaner images than a smaller sensor.

But practically, this is not an issue. 1Ds MkII has a smaller pixel size than 1Ds. So by your statement, can I assume greater noise/lower dynamic range/small color range in 1Ds MkII? No. In that case, you cannot correlate them, much less relate them at all.

 

[Zerstorer]

Oh Denial ... is a riiiiiver ... in Eeeeegypt ... :bsmilie: :bsmilie: :bsmilie:

You know, you should really stick to the discussion at hand rather than go with all this irreverent stuff.

Evading and taunting is all that you know? Next time just reply to the arguments, at least people will not think that you are just throwing smoke and hot air.

 

[dkw]

You have just made another wild claim.

If the above is true a D1 will undoubtably have more dynamic range than a 1Dmk2.

See my point about quantifying ones arguments?

This is quite unbelievable! How old is the 1D compared to the 1DMk2. Did you see the word "inherently"? Did you read the stuff I referenced? Of course with superior cicuitry and processing algorithm you can overcome photo-diode size deficiencies, but it does not change the fact that smaller photo-diodes INHERENTLY have LESS DR than bigger ones, all ELSE REMAINING the same. I quote to you again,

"Impact of Shrinking Design Rules
Forced upon machine vision by foundries
satisfying the needs of the consumer market
• leads to greater pixel resolutions in
smaller packages
• leads to smaller pixel sizes (3um-5um)
• leads to lower operating voltages (2.5V or
3.3V)
• leads to poorer dynamic range


CMOS Imaging and Application in Machine Vision
Impact to Dynamic Range
• Lower operating voltages reduce the
output voltage of the sensor amplifier
hence the electron storage capacity (full
well) of the device becomes smaller
• A reduction in full well leads to poorer
signal to noise or dynamic range because
shot noise reduces at a slower rate
• The relationship of full well to lower
operating voltages is sub-linear"

Of course the 1D has poorer DR than the Mk2, its 3 years older technology for grief's sake! I've written in previous posts that I am NOT a supporter of FF, a 20D sensor does fine for me thank you, BUT, comparing apples to apples, a larger light gathering source will ALWAYS give you more latitude to work with, that is an inescapable fact of physics, and is the only point I'm trying to make.

Of course, you guys being the argumentative folk you are , will come up with a difference nuance "but oh! you didn't qualify this, you didn't exclude that". Maybe I should call my lawyer :bsmilie: .

All right, enough, I'm sure you are all nice folk in person and I don't want to make this personal. Good luck with your photography and don't let your dynamic range get you down.

Cheers,