If I understood correctly, the DoF (all other parameters fixed) increases with the pixel surface. So to have a high sensitivity together with a deep DoF, why don't we just have big pixels (and thus large sensor surfaces, because we keep the number of pixels fixed)? What is the drawback ?
... why don't we just have big pixels (and thus large sensor surfaces, because we keep the number of pixels fixed)?
There are camera models with lower resolution sensors, but the reason they are made is for increased light sensitivity, not DOF.
If you want deeper depth of field, consider small sensors, wide-angle lenses, and small apertures. This combination is fairly common in compact cameras and cell phones.
What people usually seek is increased background blur, not DOF. People often incorrectly state this as shallow DOF (as I have in the past). However, it is possible to have same DOF, but differing amounts of background blur. See Does amount of background blur change with focal length given equal framing?
... the DoF (all other parameters fixed) increases with the pixel surface... What is the drawback ?
The problem is the relationship between DOF and parameters, like pixel or sensor size, is indirect. This is illustrated by contrary opinions. For instance, some claim large sensors produce shallow DOF, while others claim the opposite. Both claims can be correct, depending on the specific scenario and which parameters you adjust to compensate.
Consider a single-pixel sensor of any size. It pretty much does have infinite depth of field (all images are equally sharp or blurry), but it's not possible to have "all other parameters fixed". When they can be kept constant, DOF stays the same. Similarly for sensor size. When parameters are kept constant, DOF is the same.
The usual formula for Depth of Field (DOF) is:
DOF = 2 u2 N C / f2
N = aperture F-number
C = circle of confusion
u = distance to subject
f = focal length
The size of the Circle of Confusion (CoC) is chosen, arbitrarily, based on perceived sharpness. The Wikipedia CoC page states:
The common criterion for “acceptable sharpness” in the final image (e.g., print, projection screen, or electronic display) is that the blur spot be indistinguishable from a point.
The DOF formula comes from the film era. In the digital age, CoC can (but doesn't have to) be substituted with the pixel, which is the "spot" that is "indistinguishable from a point".
When pixel peeping, pixel size affects DOF by altering the size of the circle of confusion that is being considered. You can achieve the same effect by viewing the image at reduced magnification or downsizing the image, as juhist describes.
For the pedantic, let:
CoC = sensor-diagonal / [magnification * (x2 + y2)1/2]
where (x, y) is the sensor dimension in sensels ("pixels").
Those who insist that pixel size is unrelated to DOF appear to intentionally ignore the pixel-peeping scenario. If you stick with film or do not pixel peep at all, pixel size is irrelevant to you. Pixels are relevant to those who do pixel peep, even if only occasionally. Nearly all modern lens reviews involve some pixel peeping.
Sensor size affects DOF indirectly by affecting focal length, distance, and circle of confusion. When capturing the same field of view, DOF is usually increased with smaller sensors. However, it is possible to choose parameters where DOF is kept constant. See Would a 50mm lens on a Canon APS-C crop produce the exact same image as an 80mm lens on a full frame camera?
DOF Calculator for Pixel Peeping
Use the Cambridge in Colour: Depth of Field Calculator
Type in a max print dimension = pixel-width/96 inches. For a 24mp camera that produces 6000x4000 images, 6000/96 = 62.5 inches.
Leave viewing distance at 25cm, unless you're a super peeper, in which case, change it to 10cm.
Select 20/20 vision.
Select sensor size and other parameters as desired.
Although this calculator does not report the size of the circle of confusion, based on the large viewing size, it should pick something close to that of the pixel size. Note that DOF increases when sensor size (therefore pixel size) is increased.
With a 50/4 lens at about 5ft, DOF is usually reported at around 15-20cm, depending on sensor size. However, when pixel peeping, it's only about 1cm.
The DOF does not change based upon pixel size (surface area). It changes with the sensor size. This is due to a larger physical sensor area requiring less magnification for an equivalent output size. In the case of "all other parameters fixed" a different sensor size also results in a different image (different "crop factor").
DOF is not an intrinsic characteristic of an image; it all relates to enlargement/magnification (circle of confusion). If you take an image of questionable sharpness and view it small on your computer it will appear sharper. View it larger (enlarge/magnify it more) and it will look questionable. Then move away from your computer so that it appears smaller, and it will appear sharper again... that is how DOF (perceived acceptable sharpness) works.
The drawback is that if the sensor area is fixed, you have less pixels with increasing pixel size. Thus, the image resolution suffers. That's hardly what photographers want. Also, DoF depends on the pixel size only because you can't distinguish details within a pixel, but you can distinguish details between two neighboring pixels. So, you could just downscale the image and claim you have a deep DoF. An extreme of that would be to downscale the image to 640 x 480; what a huge DoF you would then have. And what low picture quality you would have!
In practice, the sensor area is fixed due to photographers having multi-thousand-dollar investments in lenses supporting only certain sensor area. A larger sensor area would require getting rid of these investments and investing even more money into larger-sensor lenses. Also, semiconductor manufacturing process has lower yield for larger sensor areas. A smaller sensor area would mean the lenses have a needlessly large image circle, and thus have too much glass, being too heavy and too expensive to do the job. Optimal lenses for small sensors are smaller and have a smaller image circle. Furthermore, the effective focal length (not the physical focal length) changes, and therefore, the uses of the lenses would change: what is a normal lens on full frame would be a short tele on crop camera.
If you increase the sensor size, you need to have a different lens. Actually, larger sensors due to the different lenses required may have a lower depth of field. It's common knowledge that if you have a crop camera and a full frame camera, to get shallow depth of field, you should select the full frame camera (but the exact details depend on what lenses you compare against each other).
So, perhaps a little counterintuitively, if you want deep DoF, use a smartphone and a lens designed for a smartphone sensor (that is sold fixed with the smartphone). I think you'll find the smartphone camera with its small pixel size and small sensor size has a deep DoF.
Larger format sizes, be it film or digital, necessitates the use of longer focal lengths lenses.
Your question revolves around the projected image cast by the lens and the fact that this image consists of countless super tiny blurred circles called circles of confusion. These are the smallest fraction of a projected image that can convey intelligence. Further, these circles must not be preserved as disks. In other words, they must be so small that an observer notices only dimensionless points of light. This is what must happen if we are to perceive an area of the image is tack sharp.
To gain depth-of-field we set our camera’s aperture to a tiny opening like f/16. If a full frame has a 50mm lens mounted, the working aperture at f/16 is 50 ÷ 16 =3.12mm. If we were using a compact digital we would mount a 30mm lens to obtain similar subject framing. If this 30mm is set to f/16, the working aperture is 30 ÷ 16 = 1.9mm. In other words, smaller formats decree smaller working aperture diameters.
It’s this simple datum --- shorter focal length lenses must be set to a smaller working diameter to achieve the same f-number. The shorter lens with its smaller working aperture spawns smaller circles of confusion -- thus the span of depth of field expands for smaller format cameras.