Has Canon fallen hopelessly behind? (cross-posted)

[Ray]

OK, so we are saying that the amount of photons reaching the sensor is the same regardless of the sensor size, but that larger sensor have a greater well capacity.

I can understand that. 

Cheers,
Bernard

No! No! No! Bernard. We're not saying that. The larger sensor always gathers more light and therefore produces less shot noise (proportional to the signal) at a specific shutter speed and specific ISO and specific aperture and specific DoF, making allowances for T-stop variations of course.

The issue is, that different designs of sensors with different quantum efficiencies (such as the CCD of MFDB) may fail to adequately suppress other sources of noise which may overshadow and make irrelevant the lower shot noise of the larger sensor.

Consider the example of the D7000 compared with the D800. The D800 sensor has over twice the area of the D7000 sensor, therefore, for any given scene of equal FoV, shot with equal T-stops, equal shutter speed and equal ISO, the D800 must gather more than twice the number of photons for a correct, ETTR exposure, assuming the pixels in both cameras have similar qualities, which they do.

The key to understanding this fact is the basic formula that describe lens aperture diameter.

The physical diameter of an aperture at a given f stop is given by the very simple formula, that I'm sure everyone reading this thread can understand, of Focal length divided by F stop.

I take a shot with my D700 with a 36mm lens at F8, and the physical diameter of the lens aperture is 36/8 = 4.5mm. That may be surprisingly small, but that's what governs the amount of light reaching the sensor for a given shutter speed (excluding considerations of glass opacity which governs T-Stop).

I shoot the same scene with a D7000 using a 24mm lens which gives me the same FoV. The precise F stop equivalent is F5.33. 24/5.33 = 4.5mm. The aperture diameter is the same in both situations, for the same DoF, except in practice one would choose either F5.6 or F5.

Now, here's the source of the confusion. One might tend to think that because the aperture diameter is the same, and because the shutter speed is the same, the same amount of light reaches the sensor and therefore perhaps shot noise is the same, (assuming always that T-stopis the same).

This initially puzzled me, until I saw the light. Whilst it's true that the same physical size of aperture, in conjunction with the same FoV, and in conjunction with the same shutter speed, and in conjunction with the same T-stop, governs the amount of light reaching the sensor, such quantity of light is equal per unit area of sensor.

The larger sensor must gather more light in the circumstances, simply because it's larger, has more units of area.

I don't know! When will I be nominated for a Nobel prize? 

[RFPhotography]

But, Ray, the shutter speed won't be the same in both cases.  If the correct exposure at f8, ISO100 is a shutter speed of 1/60 then using an aperture of f5.6 will lead to a different shutter speed to maintain the same exposure.  I understand what you're saying about trying to equate the two based on the aperture diameter.  But if you use the same shutter speed in both cases you'll have different actual exposures of the two images.  One may be properly ETTR'd.  The other won't be.

[BJL]

Bob and Ray,

But, Ray, the shutter speed won't be the same in both cases.  If the correct exposure at f8, ISO100 is a shutter speed of 1/60 then using an aperture of f5.6 will lead to a different shutter speed to maintain the same exposure.  I understand what you're saying about trying to equate the two based on the aperture diameter.  But if you use the same shutter speed in both cases you'll have different actual exposures of the two images.  One may be properly ETTR'd.

True if you use equal Exposure Index, like ISO100 in each case, as when seeking "ETTR". But if instead you use equal shutter speed along with Ray's equal [effective] aperture diameter, you gather an equal total amount of light, just spread at different intensity over the different sensor sizes. As Ray indicates, aperture ratios give a measure of the intensity of ilumination (photons per unit time per unit area), which mesh nicely with measures like a film's sensitivity and Exposure Index, which are also "per unit area" measures.

The comparisons depend on which aspects you wish to hold equal in the comparison. For example:

1. Equal composition including equal FOV and DOF along with equal shutter speed leads to equal effective aperture diameter, so that aperture rario adjusts in proportion to format size and the smw amount of light os gathered, but spread at differemt density over different sensor sizes, requiring the Exposure Index to the adjusted, in proportion to sensor area. Same light total, so skae per pixel of pixe counts are equal, but different light per unit area of sensor.

2. As above except not caring about shutter speed, one can likely use the same Exposure Index (say ISO100), which leads to the exposure duration increasing with the larger format, in proportion to sensor area. Same light gathered  per unit area, so total light (and total per pixel with equal pixel count) increased in proportion to sensor area --- a traditional scenario where a larger format has better shot noise characteristics through use of a longer exposire time.

3. Forget DOF, shoot wide open, and equip the larger format with a bigger lens: longer focal length for equal FOV, but equal minimum aperture ratio, so that the aperture diameter is larger in proportion to focal length and linear format size, and so the effective aperture area (and front lens element area?) is increased in proportion to sensor area. Then at equal shutter speed and equal Exposure Index, the larger format gathers proportionately more total light, the same amount of light per unit area. This is the scenario where so many people attribute the greater low light performance to the larger sensor alone, whereas it is hopefully clear from these details that the larger lens and thus larger aperture diameter is the main ingredient. Sensors can only gather light as fast as the lenses in front of them, and aperture diameter is the main measure of that.

[RFPhotography]

OK, yes that makes sense.  Changing the EI (i.e., ISO) will be required.  When Ray didn't indicate that I thought me might be telling us that light meters as we know them are completely useless. 

[ejmartin]

It is perfectly true that the sensor with less pixels never resolves better than the sharper sensor. What happens, though, it that the sharper (more MP) sensor starts to suffer form diffraction EARLIER than the less MP sensor of the same size. You can see that from the graph also if you interpolate the graphs a bit (lesser MP sensor sharpness peaks later than the other). The only thing that means is that you are not getting the resolution advertised unless you are careful with that aperture ring.

I think that's an error by the person that extracted the data from DxO (I just grabbed it to make the point, didn't want to spend time generating the graph myself from the original source).  If you look at the DxO data directly, both the D3 and D3x have identical center resolution at f/4 and f/5.6 on this lens (that is, for each camera, the resolution at f/4 equals that at f/5.6, with of course the D3x having more than the D3; for both, the resolution starts to fall by f/8).  So again, it is not the case that the higher MP camera starts to suffer diffraction earlier than the lower MP one.  That is a myth, pure and simple.  What the lower MP sensor does is sample the same image more coarsely, and so the resolution peak is flattened as the coarse sampling places a hard limit on what resolution can be achieved.

And of course you never get the resolution advertised; resolution is degraded every step of the way -- by lenses, AA filters, sensors, etc.  Therefore, if you want to get the most out of your lenses it makes sense to increase the sensor MP count so that that is not the limiting factor in resolution, as it is for the D3 in the above example.

[RFPhotography]

Emil, while the sensor with the higher pixel count doesn't suffer from diffraction earlier, the effects of diffraction are greater (which makes some intuitive sense since the higher resolution will show the effects more finely) than on the sensor with the lower pixel count.  The graph only goes to f16.  I'm wondering if the graph were continued out to, say, f32 whether they would meet or whether the lines would cross.

[ejmartin]

Emil, while the sensor with the higher pixel count doesn't suffer from diffraction earlier, the effects of diffraction are greater (which makes some intuitive sense since the higher resolution will show the effects more finely) than on the sensor with the lower pixel count.  The graph only goes to f16.  I'm wondering if the graph were continued out to, say, f32 whether they would meet or whether the lines would cross.

As I said, resolution is degraded by every step of the optical path.  So when you say that 'the effects of diffraction are greater with more MP', you are really saying 'the sensor is less the limiting factor in resolution with more MP'.  Isn't that a good thing?

What will happen if the lens allowed smaller aperture is that the resolution curves will meet (they will never cross) at some point.  I looked at one of the macro lenses in DxO's treasure trove of data (since the 50/1.8G only goes to f/16), and at f/32 the D3x has 31 lp/mm while the D3 has 29 lp/mm.

[RFPhotography]

OK, they'll meet but not cross.  That's fine.  It was more a question of curiosity than anything else.

And yes, I understand that the sensor isn't the only factor.  That's obvious.  I'm not suggesting that higher resolution isn't positive or isn't a good thing for photographers who need it.  Not at all. 

[hjulenissen]

People seem very afraid of "being lense-limited". Even to the point where they will refuse to buy a new, higher MP count camera. Why is that any worse than being "sensor limited"? Isnt the pragmatic issue that of being limited at all, and how restricting this limit is for your photography?

One might say that a racing car with aerodynamic tuning stop being "grip limited" and instead become "engine limited". So what as long as it goes from 4th place to winning the race in the process?

-h

[Christoph C. Feldhaim]

People seem very afraid of "being lense-limited". Even to the point where they will refuse to buy a new, higher MP count camera. Why is that any worse than being "sensor limited"? Isnt the pragmatic issue that of being limited at all, and how restricting this limit is for your photography?

One might say that a racing car with aerodynamic tuning stop being "grip limited" and instead become "engine limited". So what as long as it goes from 4th place to winning the race in the process?

-h

Most likely the reason for this discussion is to cover being mind limited ... 

[BJL]

Bob and others,

    This whole affair of complaining about one camera having worse diffraction limits or being diffraction limited "earlier" gets the bottom line of image quality completely upside down.

The reality is that the upper limit on resolution due to the effects of diffraction is exactly the same for any camera when using the same aperture, focal length and such; this is dictated by the physics of light and optics. The only thing that varies with resolution is other adverse effects on resolution due to the limitations of a particular camera, like having a sensor of lower resolution or a lens with worse aberrations.

The better way to say it is that with a sensor of lower resolution or a worse lens or whatever, resolution is "sensor/lens/camera limited" at a wider range of aperture choices, and this "camera limited resolution" sets in earlier when you start at smaller apertures (where resolution is about equal, being dominated by diffraction) and then open up (until "camera inferiority" reduces resolution more and earlier in a camera with lower resolution.)

There is no sense in which a camera offering inferior resolution wins in any practically relevant comparison of resolution!


P. S. To Christophe: yes, and many of us are predominantly "technique limited" anyway, so I am not sure why I spend so much time thinking about this! Sometimes it is a bit like a bunch of unfit old codgers debating whether to climb Everest or K2.

[Ray]

But, Ray, the shutter speed won't be the same in both cases.  If the correct exposure at f8, ISO100 is a shutter speed of 1/60 then using an aperture of f5.6 will lead to a different shutter speed to maintain the same exposure.  I understand what you're saying about trying to equate the two based on the aperture diameter.  But if you use the same shutter speed in both cases you'll have different actual exposures of the two images.  One may be properly ETTR'd.  The other won't be.

Absolutely right! I'm glad someone's alert. What was I thinking of! Have I had too many glasses of wine? I'm never going to get my Nobel prize at this rate. I'll have to rethink this. 

Whatever the aperture diameter and shutter speed, if they are both the same, the same amount of light passes through the lens, provided the FoV is the same (and the T-stop of course). The FoV relates to the source of the light gathering capacity.

If the lens 'field of view' is the same, and the over all, or average light intensity is the same, and the lens aperture diameter is the same, then for any given exposure the same number of photons should pass through the lens (discounting shot noise variations).

Therefore, a 24mm lens at F5.6 (or more precisely F5.33) should let pass the same amount of light as a 36mm lens at F8, using the same shutter speed.

So why does the wider angle lens with the smaller F/stop number (but same physical aperture diameter) require a faster shutter speed? Good question.

Answer: Because the same light-gathering capacity, wshich is directly related to the same field of view, is directed to a smaller sensor, in the case of the cropped format. The smaller sensor, if the FoV of the scene and the light intensity is the same, would have to receive the same amount of light as the FF sensor used with the 36mm lens at F8, if the exposure were the same. In order for it to do that without overexposure, the ISO would have to be half that of the full-frame sensor, and its sensels more efficient and deeper.

This is not usually the case, so the smaller sensor usually receives less light in relation to the same amount of scene detail, ie. the same field.

Crikey! That was a good glass of Chardonnay. 

[Thomas Krüger]

Complicated discussion. On my shopping list for 2012 is a lightweight mirrorless like the Nex-7 and a high-res DSLR body like the Nikon D800 with a pair of prime lenses, maybe also from Zeiss with excellent mechanics. Having still the old 5D I can't find something interesting for me in the actual product line of Canon. But perhaps this will change with the upcoming Photokina, who knows.

[Petrus]

 So again, it is not the case that the higher MP camera starts to suffer diffraction earlier than the lower MP one.  

Yes. Correct. The lens resolution peaks at a certain place (diffraction wins over the quality gains from excluding the lens edges). All sensors give the best result at that point. No matter what the pixel count is. Diffraction just shaves off the "extra" resolution of the bigger MP sensor a bit faster, maybe, but resolution stays above the lesser sensors all the way to the smallest apertures.

[BJL]

Diffraction just shaves off the "extra" resolution of the bigger MP sensor a bit faster, maybe, but resolution stays above the lesser sensors all the way to the smallest apertures.

It is of far more practical relevance to say that sensor limitations "shave of resolution faster" as you open up from small apertures!
The "diffraction limited" view is like complaining that a motorcycle is more "legal speed limit limited" than a moped, making a performance virtue out of the fact that the moped can never go as fast as the legal speed limit.

P. S. The subject line that I created long ago is now a bit off-topic, but no one gave a straight answer to that question! So, if 35MP, 40MP and more have their place, why not in 35mm format, given the very favorable per-pixel comparisons between the D800/D800E and MF options of similar pixel count?

[Petrus]

P. S. The subject line that I created long ago is now a bit off-topic, but no one gave a straight answer to that question! So, if 35MP, 40MP and more have their place, why not in 35mm format, given the very favorable per-pixel comparisons between the D800/D800E and MF options of similar pixel count?

High MP 35mm format cameras seem to earn their rightful place, as at least D800 seems to be better than expected what comes to low light performance. Still I can see reasons why workhorse press & sports photographer cameras might stay below 20 MP mark: they still have a slight edge in the dark, and the smaller (small?) files need less processing, thus faster frame rates can be offered. I will have it both ways, as later this year I will be getting both D800e and D4 with 4 fast zooms from 14 to 400mm (replacing 5DII and 1DII and a sack full of Canon glass). With X-Pro1 already in my bag (easily equals 5DII in resolution) the future looks bright, dark, and every shade in between!

And I try not to use f-stops smaller than f:6.3... Promise.

[RFPhotography]

Bob and others,

    This whole affair of complaining about one camera having worse diffraction limits or being diffraction limited "earlier" gets the bottom line of image quality completely upside down.

For the record, I wasn't complaining about any of it.

There is no sense in which a camera offering inferior resolution wins in any practically relevant comparison of resolution!

No argument there.  I was just trying to clear up a couple things.

[RFPhotography]

So why does the wider angle lens with the smaller F/stop number (but same physical aperture diameter) require a faster shutter speed? Good question.

For the record, I didn't ask the question so you're patting yourself on the back. 

Answer: Because the same light-gathering capacity, wshich is directly related to the same field of view, is directed to a smaller sensor, in the case of the cropped format. The smaller sensor, if the FoV of the scene and the light intensity is the same, would have to receive the same amount of light as the FF sensor used with the 36mm lens at F8, if the exposure were the same. In order for it to do that without overexposure, the ISO would have to be half that of the full-frame sensor, and its sensels more efficient and deeper.

Which is effectively what Emil noted above.  And that was the point of my question, to try to add that extra level of clarity.

[BJL]

For the record, I wasn't complaining about any of it.

Indeed: my comment referred to other people (and mostly in other threads, and even at other sites) who keep claiming that having a wider range of f-stops at which the resolution is not "sensor limited" is a bad thing!

[Ray]

Now that I'm flush with cash and have a few banknotes in my wallet, I can confirm that those broad bands sloping down from left to right, on each side of David Unaipon's face in the 40D shot of the $50 bill at F8, are indeed moire or aliasing artifacts.

If I'd used the D800E for this shot, using a similar quality of lens, I'd expect to see more obvious examples of moire, but at F16, probably not.

One issue may be, if diffraction at F16 removes the need for an AA filter, is resolution still slightly compromised at F16 if the sensor also has an AA filter? Are the two blurring effects of diffraction and AA filter additive?