Sensor density and diffraction

Have questions about the equipment used for macro- or micro- photography? Post those questions in this forum.

Moderators: Pau, rjlittlefield, ChrisR, Chris S.

gmazza
Posts: 566
Joined: Sat Dec 12, 2009 4:03 pm
Location: Rio Grande do Sul; Brazil; 29°S 51°W

Sensor density and diffraction

Post by gmazza »

I'm stuck with this.

With the increase in megapixels of the industry I'm wondering if a older low MP camera would be better than a new one, for example, a "classic" Canon 1D with 4 MP 1.3 crop sensor have a pixel size of about 11.5 µm wich looks awesome to avoid diffraction. A, not so old, Canon 1D MK II pixel size of about 8.3, wich looks a midrange compromisse.

Actual models as the Canon 5D MK II have pixel size of about 6.5 µm and the canon 7D pixel size of about 4.3 µm (aproximatte numbers)

Opposite to this idea are the newer tecnologies used in the modern sensor with less noise and better color.

With the diffraction calculators even the 11.5 µm sensor become diffraction limited after f/16 (and my handheld shoots use an effective of about f/40 to f/60) there is possibly a difference even after the limit of diffraction was surpassed ?

With most of my photos being single frame handheld shoots diffraction is afecting all of them, and all info that I read about it point to sensor size/density.

And finally, comparing the 4 MP of 1D MKI with the 8 MP of 1D MK II (being the two 1.3 crop factor sensors), even the 8 MP being diffraction limited reducing the image size of it to the same delivered by the 4MP 1D classic would clear the difference ?
Last edited by gmazza on Mon May 17, 2010 1:07 pm, edited 1 time in total.
Gustavo Mazzarollo

Portfolio

http://www.gmazza.com

Charles Krebs
Posts: 5865
Joined: Tue Aug 01, 2006 8:02 pm
Location: Issaquah, WA USA
Contact:

Post by Charles Krebs »

Gustavo,

What is happening with a higher pixel camera is that (at the same lens settings) you are "placing" more pixels under the Airy disc (blur disk). If the sensor sizes are the same, you really gain nothing as far as diffraction effect is concerned. The actual amount of diffraction at the sensor plane depends on the aperture and magnification. The sensor "density" (pixel size) determines at what point we start to notice that the given amount of "diffraction" is causing a noticeable decline in resolution. So in a simplified sense... the lower resolution sensor is does not cause less diffraction, it's just that it does not resolve high enough that the diffraction blur becomes large enough to be the limiting factor until we have "stopped down" past the point where the higher resolving sensor has already "seen" it. If the sensor sizes are the same, the larger pixels simply need to encounter a larger "blur" before it can even be considered a problem, and thus it does not become "diffraction limited" until after (smaller aperture) the higher density sensor.

However the "less dense" sensor will be giving up potential resolution in the cases where the optics are providing an image with smaller "airy disks".

So when you look at the pixel size and think:
looks awesome to avoid diffraction.
It doesn't avoid diffraction. The real reason it sounds better (diffraction limited at a smaller aperture) is because it can't resolve high enough that the diffraction that is there interfers with it's lower resolution capability.
Here in the macro world we're primarily concerned with DOF. Diffraction limited depth of field is constant for all sensor sizes. (A larger sensor is "enlarged" less, but it requires more magnification to fill the frame with subject size.)

But.... (always a but... :wink: )

There are some things in favor of larger pixels (with current technology) that are not diffraction related, and those would be "noise" and dynamic range issues. These should be factored in as well to determine what might be "best" in any given situation.
Last edited by Charles Krebs on Mon May 17, 2010 11:48 am, edited 2 times in total.

g4lab
Posts: 1434
Joined: Fri May 23, 2008 11:07 am

Post by g4lab »

Gustavo you have asked a question that I too have been wondering about.
The number of pixels keeps going up. Fortunately there seems to be some trend also to make the sensor size bigger.

But in spectroscopy bigger pixels are usually better. They have a better signal to noise ratio than smaller pixels. But also less resolution for a given amount of dispersion by the dispersing element.

I have some recollection of this being discussed previously. Perhaps Rik remembers and can give us a link or can give us one of his patented superb clarification lectures on the topic. :D

mgoodm3
Posts: 273
Joined: Mon Sep 08, 2008 8:50 am
Location: Southern OR

Post by mgoodm3 »

Keep in mind that the Airy disc for most images above 1:1 is significantly larger than most detector sites, so your detector will not be working at its optimal effeiciency anyway. I think that the Airy disc is about 20um at f8 and 1:1, that's 3x larger than the detector on a D200. Image quality is predominantly related to the lens/aperture and not the detector and it just gets worse as the magnifcation goes higher.

Adding megapixels doesn't help a whole lot unless you can increase the detector size. You may get a little better image quality with more, smaller detector sites through oversampling, but that difference should be pretty small and possibly balance by the increased noise with smaller sites.

PaulFurman
Posts: 595
Joined: Sat Oct 24, 2009 3:14 pm
Location: SF, CA, USA
Contact:

Post by PaulFurman »

Newer cameras generally perform better regardless of pixel size (within reasonable limits). For your situation it sounds like it would make sense to just shoot a smaller megapixel setting on the camera. Canon even has a reduced size raw format.

Note that dxomark numbers show the 5DII with better high ISO performance than the old 5D even though the megapixels moved from 12 to 21.

Newer cameras also have nice features like larger LCD display, video, live view, faster write times, etc.

rjlittlefield
Site Admin
Posts: 21397
Joined: Tue Aug 01, 2006 8:34 am
Location: Richland, Washington State, USA
Contact:

Post by rjlittlefield »

g4lab wrote:Perhaps Rik remembers and can give us a link
All the discussions I remember are so wrapped up with other issues that I think they would not be helpful. Let me try summarizing from scratch.

Short answers:

1. The newer camera with more pixels is probably better, for all the reasons that Paul and Charles outline: improved technology and potential for higher resolution if the optics provide it.

2. Larger pixels give less pixel noise, but not necessarily less image noise. The differences in image noise are due to secondary issues and are much smaller than you might think from casual pixel-peeping.

Discussion:

As Charles notes, having smaller pixels on the same size sensor does not mean that diffraction blur get worse. It just means you have more pixels in which to see the blur. In essence, the potential for higher resolution provided by smaller pixels becomes wasted when diffraction is the limit.

Noise is a trickier problem.

To a first approximation, camera sensors are just photon counters. The arrival of photons is a random process, so that even under constant illumination there will be variation in the number of photons counted by each pixel. This sampling variation is the main cause of pixel noise.

It is a property of random sampling that relative variation gets smaller as counts get larger.

Larger pixels can count more photons. So, if you look at the noise on a per-pixel basis, larger pixels give less noise.

But there is a flaw in that approach. Under most circumstances, the viewer of an image cannot see the camera's pixels. Instead, each small area perceived by the viewer is an average of several pixels from the camera. When the camera's pixels are smaller, more of them get included in the average for any particular piece of the image. This averaging reduces the noise.

When you slog through the math, it turns out that the increase in pixel noise due to smaller counts with smaller pixels is exactly canceled by the decrease in noise due to averaging across more of those pixels in the same image area. In other words, there is no difference in image noise, although smaller pixels have more pixel noise.

There are, of course, some assumptions built into this analysis. The two biggest assumptions are that

1) electronic noise is not significant (that is, pixel noise is due entirely to sampling uncertainty), and

2) small and large pixels have the same capture efficiency (a combination of percent active area and quantum efficiency).

Those assumptions are not completely true. On the one hand, larger pixels may allow a higher percentage of active area and thus give a higher capture efficiency, while on the other hand I've read that smaller pixels may suffer less from sense amplifier noise due to the smaller well giving higher voltage per photon.

Ultimately, I think that noise comes down to "try it and see". There are too many unknowns to predict accurately from theory. I have seen reports of head-to-head comparisons with every possible outcome.

--Rik

gmazza
Posts: 566
Joined: Sat Dec 12, 2009 4:03 pm
Location: Rio Grande do Sul; Brazil; 29°S 51°W

Post by gmazza »

Thanks everyone for valuable thoughts, I really performed searches to check if was discussed before, sometimes into other discussions, but never as "the topic" so I'm happy more forumattes were wondering about this.
Have seen it discussed in astrophotography forums (another form of photo affected by diffraction and noise) were the overall conclusion was that a 4 MP 1.3 crop factor sensor is not better than new technology. But never seen it discussed in macro context.
Charles Krebs wrote: The actual amount of diffraction at the sensor plane depends on the aperture and magnification.
It doesn't avoid diffraction. The real reason it sounds better (diffraction limited at a smaller aperture) is because it can't resolve high enough that the diffraction that is there interfers with it's lower resolution capability.
Here in the macro world we're primarily concerned with DOF. Diffraction limited depth of field is constant for all sensor sizes. (A larger sensor is "enlarged" less, but it requires more magnification to fill the frame with subject size.)
I understand and agree with this, in face of the information I could conclude that for quality images of very small subjects there are two possible aproaches without breaking the laws of physics:

1. Be more tolerant about cropping some images

2. Improve handheld focus stack skills
rjlittlefield wrote: As Charles notes, having smaller pixels on the same size sensor does not mean that diffraction blur get worse. It just means you have more pixels in which to see the blur. In essence, the potential for higher resolution provided by smaller pixels becomes wasted when diffraction is the limit.
This is compative with what I'm observing with my 5:1 MP-E shoots at smaller spertures, they only looks good with reduced size. Sometimes I think a 4:1 shoot plus crop would work better than a 5:1 downsized, but diffraction is not the only factor into play, working at lesser magnification also reduce light losses and flash times (less motion blur, better linnear gamma profile, probably less noise depending on exposure choices) Lesser magnification also increases the DOF allowing work with larger aperture for the same dof, increasing even more the light.

For example the Canon MP-E manual describe a DOF of 0.175mm @4x and f/8 (effective aperture of f/40) and 0.190 @5x and f/11 (effective aperture of f/66). Considering the 0.015mm difference minimal is a big compromisse for a 4mm frame instead of a 5mm one (and in a FF camera would be a 8.75mm frame @ 4x and a 7mm frame @ 5x with the same DOF)
I think for some subjects the trade off could be positive to cropping.

Two real life examples:

This is a crop from 4752 x 3168 to 3599 x 2879
It's a crop but even with the possibility of making a 24x36 Inch print
Image
Canon EOS REBEL T1i (15MP 1.6 crop factor sensor)
Canon 100mm macro f/2.8 with Raynox MSN-202 in front of it (wich give 4:1)
f/16 (effective about f/32) | iso200 | 1/125s

Of course these Robber Flies have awesome eyes and my ambition is to fill the complete image with them, so this morning I got this dew covered frame from a steadier one, have obvious focus problems related to small DOF.

Resolution 4752 x 3168
Image
Canon EOS REBEL T1i (15MP 1.6 crop factor sensor)
Canon MP-E 65mm f/2.8 @ 4x with Raynox MSN-202 in front of it (wich give 8:1)
f/8 (effective about f/40) | iso200 | 1/125s
Gustavo Mazzarollo

Portfolio

http://www.gmazza.com

Charles Krebs
Posts: 5865
Joined: Tue Aug 01, 2006 8:02 pm
Location: Issaquah, WA USA
Contact:

Post by Charles Krebs »

Gustavo,
Have seen it discussed in astrophotography forums (another form of photo affected by diffraction and noise) were the overall conclusion was that a 4 MP 1.3 crop factor sensor is not better than new technology.
Frankly, if the discussion were limited to photomicrography at higher magnifications my conclusion would be more in line in line with that. Primarily because diffraction levels are so high that even the finest optics can be "handled" with fairly modest (by current standards) pixel counts. My old 6Mp Canon 10D was more than adequate for work with a 20/0.40 and higher magnification objectives (when used with a 1.67X photo-eyepiece). But with a 4/0.16 Plan Apo (or a 10/0.40) it did not have enough pixels to get all the detail being produced. So with the microscope, after a certain point, all the extra pixels are not needed.

But we generally use a wide variety of optics and magnifications, so it is nice to have the pixels there when using a set-up that can make full use of them.

Post Reply Previous topicNext topic