Pages: [1]   Go Down

Author Topic: PIXEL DENSITY  (Read 14710 times)

EAD

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 70
PIXEL DENSITY
« on: September 16, 2005, 12:39:25 pm »

I must admit I am not into the thechnicall part of this hobby/job very much. I love my photography, and from time to time I even make some money out of it, but I´m nowhere close to some of the guys here like Jonathan, JBL, Jack Flesher, etc..(all my admiration for them)

One thing I need to understand is (a lot of missinformation around..) how exactly does pixel density affect the capture, the lenses and all the parts involved in taking a picture. I know my 20D has the highest pixel density of any current Canon DSLR. Apart from the signal to noise ratio (which, whatever the tradeoffs seems to be pretty under control by the wise guys in Canon), what is the difference between two chips with same pixel count but different size /density. What are the advantages or disadvantages of each (if any..).

I thought it was the gross pixel count that was putting the lenses in a difficult situation regarding resolution, but I´ve read somewhere that a high density chip like the one on 20D is very demanding with the lenses...I´m a bit lost :cool:

please throw some light over my ignorance...  

Thanks.

Erik
Logged

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
PIXEL DENSITY
« Reply #1 on: September 16, 2005, 06:47:24 pm »

Gross pixel count refers to the total number of pixels on the sensor, whatever the size of the sensor. Pixel density refers to the number of pixels per fixed area, such as square millimetre. If 2 different sized sensors have the same gross pixel count, such as the Sony F828 P&S and the Canon 20D DSLR (both 8MP), then the smaller sensor will have the greater pixel density; in this case the F828.

The greater the pixel density, the higher the resolving capacity of the sensor (in terms of lines per mm) and the greater the demand on the lens. A standard 35mm lens would not be good enough for many a small P&S camera. The sensors of such cameras are so small (about 1/16th the area of full frame 35mm), the lenses need to be very, very sharp, and they can be because they are so small and cheap to manufacture.
Logged

jani

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1624
    • Øyet
PIXEL DENSITY
« Reply #2 on: September 16, 2005, 06:57:38 pm »

I'll try to respond, but only partially.

If all else is equal, a higher pixel density means that the sensor is more demanding of the lenses.

However, it can be really hard to say if all else really is equal.

Also, if all else is equal, while e.g. a 20D is fairly nasty towards lenses in some ways, it's also very nice in that it only uses the center of the lenses. Since the edges of the lenses -- especially noticeable in wide angle lenses -- tend to show more softness, less contrast etc., the 20D gets to use more of the better part of the lens than e.g. a 5D.

Did that make sense?
Logged
Jan

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6600
PIXEL DENSITY
« Reply #3 on: September 16, 2005, 06:59:48 pm »

Quote
The sensors of such cameras are so small (about 1/16th the area of full frame 35mm), the lenses need to be very, very sharp, and they can be because they are so small and cheap to manufacture.
Ray,

   one other thing I learnt recently about tiny format digital cameras: the lenses for non-SLR digital cameras take great advantage of not having the reflex mirror, using designs where the rear lens elements are very close to the sensor. That is they have extremely short "back focus distance", but still high "exit pupil" ans so are near-telecentric, the way that most electronic sensors prefer their lenses to be. (Unfortunately, this must apparently come at the cost of using the inferior contrast detection style of auto-focus.)
Logged

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
PIXEL DENSITY
« Reply #4 on: September 16, 2005, 08:07:05 pm »

Quote
(Unfortunately, this must apparently come at the cost of using the inferior contrast detection style of auto-focus.)
BJL,
I've not delved into this issue, but I've realised that accurate auto-focussing in general comes at a price. I was disappointed the 20D could not autofocus at f8.

It'll be interesting to see just how good the Sony DSC-R1's autofocussing is.

A contrast detection system implies the lens must be very sharp for good focussing.
Logged

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6600
PIXEL DENSITY
« Reply #5 on: September 16, 2005, 08:22:21 pm »

Almost no camera can AF with lenses of maximum f-stop higher than about f/5.6 (a few high end models can though). How did you discover this problem? Using a 1.4x TC on your 100-400 f/4-5.6 lens?

I am not sure why you say the Sony R1's use of contrast detection AF implies anything about the good quality of the lens: contrast detection AF is simply what it and all non-SLR digital cameras are stuck with, to their detriment compared to SLR's. I hope you are not trying to turn a clear disadvantage of the R1 into an advantage, in you new enthusiasm for that camera as a further enemy of the mainstream formats of DSLR.
Logged

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
PIXEL DENSITY
« Reply #6 on: September 16, 2005, 08:46:02 pm »

Quote
I am not sure why you say the Sony R1's use of contrast detection AF implies anything about the good quality of the lens: contrast detection AF is simply what it and all non-SLR digital cameras are stuck with, to their detriment compared to SLR's. I hope you are not trying to turn a clear disadvantage of the R1 into an advantage, in you new enthusiasm for that camera as a further enemy of the mainstream formats of DSLR.
BJL,
Could I be so devious? Contrast is related to sharpness. It follows, if you have a contrast detection autofocussing system in combination with a lousy lens, you get sub-optimal focussing. If you have a sharp, and therefore contrasty lens, (within the relevant resolution parameters), you must get better autofocussing.
Logged

EAD

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 70
PIXEL DENSITY
« Reply #7 on: September 16, 2005, 09:13:43 pm »

Thanks for your responses..

So, then, should I understand that the probable next generation aps sized chip with 12-13 MP should outresolve the 5D for instance(or 1DS2)? Will it have, then, an advantage over a full frame camera with the same pixel count? If so, why all the excitement aboutr FF (apart from the wide angle)

Noise levels well under control, with the existence of the 10-22 (which seems to be a very good lense), and the clear advantage on the tele side, it seems to me everything in an APS-sized camera are advantages (unless of course you want a tank-built 1series..)

Am I missing something?

Thanks.
Logged

Jack Flesher

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2592
    • www.getdpi.com
PIXEL DENSITY
« Reply #8 on: September 16, 2005, 09:40:55 pm »

There are several issues here, and it would take several pages to properly explain them all...

But in digest form:

1) Higer pixel density is more demanding on lenses becuase by default it offers higher resolution to examine them at (as already explained).

2) Larger sensors are more demanding because they capture visual data near the outer edges of the lens' "Image Circle" and any given lens' performance will drop off as you move away from the central optical axis.

3) Smaller pixels -- the direct product of higher pixel density -- generate more "noise." Noise in turn can affect one's ability to ascertain resolution.

4) Conversely, larger pixels (lower pixel density) are less noisy or "cleaner" and hence can improve our ability to ascertain resolution.

5) Finally, raw pixel count is probably what matters most assuming the lens in front of the sensor can deliver resolution that exceeds the sensor's pixel pitch.  For example, the 8MP 1D2 needs (roughly) a 38mm lens mounted on it to generate a "normal" full-frame 50mm field of view.  By contrast the 8MP 20D only needs 30...  BUT using those focals on those two cameras from the same shooting position at the same ISO and same exposures will yield the SAME information on their 8MP sensors -- and the resultant images will look....  Different.  But not from a resolution standpoint.

6) Different how, you might ask. Because by default, the 28mm lens from the same shooting position as the 38mm lens yields greater DOF at the sensor if both lenses are set to the same aperture.  (I am NOT going to open up this can of worms and delve into the math, but trust me that regardless of the fact a few may attempt to argue this point differently, the above statement is true.)  

In conclusion, it is MY opinion that just like with film, when all else is equal there is no replacement for physical sensor acerage...  IOW, I'd take 11MP full frame over 11MP from a 1.5x sensor 8 days a week -- as long as my lenses were good enough for the full-frame camera.

Cheers,
Logged
Jack
[url=http://forum.getdpi.com/forum/

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
PIXEL DENSITY
« Reply #9 on: September 17, 2005, 12:10:46 am »

Quote
I'd take 11MP full frame over 11MP from a 1.5x sensor 8 days a week -- as long as my lenses were good enough for the full-frame came
Jack,
Are you implying that Nikon lenses are better than Canon lenses? (D2x versus 1Ds).
Logged

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
PIXEL DENSITY
« Reply #10 on: September 17, 2005, 12:13:34 am »

Quote
How did you discover this problem? Using a 1.4x TC on your 100-400 f/4-5.6 lens?
BJL,
I missed this point. Yes. Exactly. Is there a flaw in this deduction?
Logged

Jack Flesher

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2592
    • www.getdpi.com
PIXEL DENSITY
« Reply #11 on: September 17, 2005, 08:48:45 pm »

Quote
Quote
I'd take 11MP full frame over 11MP from a 1.5x sensor 8 days a week -- as long as my lenses were good enough for the full-frame came
Jack,
Are you implying that Nikon lenses are better than Canon lenses? (D2x versus 1Ds).

No.  I'm implying I'd buy a 5D before I bought a D2x IF I had good enough lenses -- and fortunately I do!  As I stated above, almost any full-frame lens is going to perform better on a crop sensor...
Logged
Jack
[url=http://forum.getdpi.com/forum/

lester_wareham

  • Full Member
  • ***
  • Offline Offline
  • Posts: 116
    • http://www.zen20934.zen.co.uk/
PIXEL DENSITY
« Reply #12 on: September 19, 2005, 07:36:05 am »

Quote
So, then, should I understand that the probable next generation aps sized chip with 12-13 MP should outresolve the 5D for instance(or 1DS2)? Will it have, then, an advantage over a full frame camera with the same pixel count? If so, why all the excitement aboutr FF (apart from the wide angle)
The reason for excitment about full frame is (or should be) for those using lenses in the ultra-wide to normal range. The full frame sensor will permit a 1.6X longer lens than for the APS-C sensor.

Very short lenses for SLRs are difficult to design and have limited performance so there is a major advanatge on two counts:

1) Lens quality (sharpness and other factors) will be better due to the less demanding focal lengths.

2) Because of the larger image better microcontrast will result just because fine detail will be at 30lp/mm instead of 48lp/mm.

If you are using long lenses then pixel density is important, but given the same pixel density and anti-alias filter characteristics a 16Mp full frame cropped will be the same same as a 8Mp APS-C. The key point is the anti-alias filter which is largely an unknown if you don't work for Canon.

The 5D this is not that exciting for wildlife photographers, except perhaps for macro work where APS-C has an advantage but a diminishing one as magnification increases.

However, for landscape folk it is very exciting.
Logged

jd1566

  • Full Member
  • ***
  • Offline Offline
  • Posts: 149
PIXEL DENSITY
« Reply #13 on: September 19, 2005, 07:58:14 am »

One item that is often overlooked when comparing full frame vs. crop factor cameras is that the lenses used remain the same, just that you are using more or less of the lens as the case may be.  I have recently used the 10-22mm lens on my 20D and been particularly disappointed in the visual effect that the wide end gives to my pictures. NOT comparable to my 17-40 on a film body (full frame negative).  This is because a 10mm lens distorts more than a 17mm. For this reason I am chosing the full frame alternative now offered by Canon.
Another point against the 20D is that the banding issue really comes to a fore in low light situations, or in underexposed photos.  This was not present on my 10D.. I am ASSUMING that the 5D (due to larger pixel size) will not have as much of a problem in this regard, though some of the upcoming reviews will certainly reveal more about it's performance that we already know.  However standard reviews do not test this type of performance by deliberately underexposing photos to see how they come out.  Anyone out there with a pre-production 5D want to try this? (Hint hint Michael!) :-)
Logged
B&W photographer - Still lifes, Portrait

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6600
PIXEL DENSITY
« Reply #14 on: September 19, 2005, 01:47:04 pm »

Ray, perhaps you are missing my main point: no matter how much Sony improves the performance of its contrast detection AF system by improving the lenses, it will not be as good as a phase difference AF system. One great advantage of phase difference AF is that it measures in which direction the lens is out of focus, so the camera knows which way to adjust the focus. Contrast difference does not give this directional information, so that the AF has a fifty-fifty chance of trying the wrong direction first, leading to more "hunting" and slower AF.

What is more, DSLR's can easily be used with lenses of larger apertures than the R1's, giving another advantage in AF performance.
Logged

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6600
PIXEL DENSITY
« Reply #15 on: September 19, 2005, 02:07:18 pm »

Quote
... when comparing full frame vs. crop factor cameras is that the lenses used remain the same, just that you are using more or less of the lens as the case may be.
...
I have recently used the 10-22mm lens on my 20D and been particularly disappointed in the visual effect that the wide end gives to my pictures. NOT comparable to my 17-40 on a film body
The first statement in nonsense, refuted by what follows: for FOV from wide angle up to moderate telephoto, smaller format DSLR's usually do NOT use the same lenses: they use lenses with designs optimized for the angular coverage desired over the smaller image circle: Nikon DX, Olympus Four Thirds, Pentax DA, Canon EF-S, Konica-Minolta DT, etc.

Also, it is rather pointless comparing Canon's amateur level 10-22 EF-S variable aperture ratio zoom to the professional level 17-40 L constant aperture ratio lens. Canon seems to have a clear goal to hold the quality of EF-S down below full professional level, so as to sustain the differentiation of its 24x36mm format DSLR being the only professional level option from Canon.

But not the only professional level option from other makers. Try instead comparing to a smaller format DSLR lens from a maker who aspires to making professional level products in the smaller DSLR formats: a Nikon 12-24 f/4 DX, Olympus 11-22 f/2.8-3.5 or Olympus 7-14 f/4.


By the way, it is funny/sad reading on Canon's new CMOS website that 24x36 format is better partly because smaller formats lack fish-eye lenses. There is Nikon DX 180º fish-eye already, and a Four Thirds 180º fish-eye has been announced. That whole site seems to be aimed at readers who consider only Canon's options.
Logged

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6600
PIXEL DENSITY
« Reply #16 on: September 19, 2005, 02:21:04 pm »

Quote
Quote
How did you discover this problem? Using a 1.4x TC on your 100-400 f/4-5.6 lens?
BJL,
I missed this point. Yes. Exactly. Is there a flaw in this deduction?
No, you are absolutely right: AF fails when the maximum aperture ratio get too small, and with all but the best SLR's, the limit seems to be about f/5.6 or a bit slower.

I asked to confirm that you are interested in telephoto reach beyond 400mm with your 20D. (Me too: A TC and/or sensor with smaller pixel spacing is second on my wish list after a macro lens.)  That alone seems a good reason to hold out for your next DSLR to have pixel spacing no larger than the 6.4 microns of the 20D. So I presume you are waiting on an affordable Canon 24x36mm format DSLR with more than 20MP.


P. S. Apparently it is f-stop that counts for AF operation, regardless of aperture diameter or format. That seems to be why f/5.6 is such a common minimum f-stop at the long end of slower zoom lenses.
Logged

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6600
PIXEL DENSITY
« Reply #17 on: September 19, 2005, 04:22:24 pm »

Quote
The reason for excitment about full frame is (or should be) for those using lenses in the ultra-wide to normal range. The full frame sensor will permit a 1.6X longer lens than for the APS-C sensor.

Very short lenses for SLRs are difficult to design and have limited performance so there is a major advanatge on two counts:

1) Lens quality (sharpness and other factors) will be better due to the less demanding focal lengths.

2) Because of the larger image better microcontrast will result just because fine detail will be at 30lp/mm instead of 48lp/mm.
...

However, for landscape folk it is very exciting.
I agree with the last point; larger formats always have been and probably always will have an edge for landscape photography that pushes the limits of fine image detail, fine tonal gradations and such, usually in normal to wide angle FOV.

But your argument about it being harder to design wide angle lenses for smaller formats is not persuasive; it seems to make the common mistake of confusing the difficulties of  getting wider angular FOV (shorter focal lengths in the same format) with of "shorter focal lengths with the same FOV and a smaller image circle", which is what matters in comparing th wide angle ability of different formats.

With wide angle lenses, increasing angular FOV is the main optical difficulty.


Here are two crude options for downsizing wide-angle lens designs to get designs giving the same angular FOV a a shorter focal length for a smaller format.

1. Scale down all dimensions of an existing wide angle lens design by the format factor (1.5x, 1.6x. 2x etc.)

2. Combine an existing wide angle lens design with a built-in custom designed focal reducer (the opposite of a teleconerter).

Option 1 reduces the focal length but gives the same angular FOV over a smaller image circle and the same aperture ratios. For example, with 1.4x it turns a 28-70 f/2.8 into a 20-50 f/2.8. This also shrinks the length scale of abberations arising from the lens design, so increasing the computed resolution and MTF behavior measured in lp/mm in proportion to the format change. That means that in terms of "line pairs per picture height", the performance will be about the same.


Option 2 of a focal reducer of 1.5x, 1.6x. 2x etc. shrinks the image formed by the lens by the indicated factor, giving an image of the same angular coverage and the same aperture *diameter* and hence proportionately reduced aperrture ratios.  With 1.4x, it turns a 28-70 f/2.8 into a 20-50 f/2.0!.

This reduction also shrinks the size scale of all abberations and resolution limits of the original lens in the same proportion as the image itself, and so shrinks the resolution length scale of the original image, automatically increasing the resolution in the sense of lp/mm, but leaving it about the same in terms of "lp per picture height", which is what matters in the end.

The decrease in aperture ratio makes the lens "faster" by enough to offset the lower ISO speed limits of a smaller format, and to give th sme DOF wide open.

"Reduction" also increases the exit pupil, making the lens more telecentric, which sensors like.


For some practical examples, compare the MTF charts for Olympus Four Thirds format lenses at 60lpmm and 20lp/mm to those for Canon 35mm format lenses at 30lp/mm and 10lp/mm (roughly equal lp per picture height, so a fair comparison). The Four Thirds lenses match up fairly well, showing that downsizing can hold up in terms of resolution needs.

In particular, the Olympus 7-14 f/4 for Four Thirds is maybe the state of the art for good performance at very wide angular FOV in smaller formats, and the forthcoming constant f/2 FourThirds zooms might well have used a mixture of both the above strategies, starting from constant f/2.8 designs for traditional 24x36mm format.
Logged

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6600
PIXEL DENSITY
« Reply #18 on: September 19, 2005, 04:58:16 pm »

Quote
3) Smaller pixels -- the direct product of higher pixel density -- generate more "noise."

6) ... Because by default, the 28mm lens from the same shooting position as the 38mm lens yields greater DOF at the sensor if both lenses are set to the same aperture.
Jack,

   you have some good points, but the two I quote are at best ambiguous, and are particualry dubious on yout rterms of comparison with equal pixel counts, whc I accept asa god way to compare.

As is so often done, you are silently assuming that the smaller sensor and smaller pixels will be used with a lens of not only smaller focal length but also of smaller aperture size (aperture diameter), in proportion to the pixel and sensor size. This assumption leads to the need to use the same high ISO speed setting when a given high shutter speed is needed. Thus the smaller pixels and format are asssumed to use a probably smaller, lighter, cheaper lens, leading to more DOF and less light gathered to each photosite, lwoering S/N ratio.

Of course, using a smaller, lighter, cheaper lens with a smaller aperture diameter has exactly the same effects even if you use the same size pixels and sensors, so I do not know why people keep blaming this on the sensors or pixels instead of the uneven choice of lenses hidden in such comparisons.

At most, sometimes a smaller format and focal length limits one to a smaller aperture size, and then one is truly limited to lower speeds, higher noise and/or more DOF along with likely weight and cost savings.


But in the telephoto regime at least, size, weight, cost, and DOF needs tend to set roughly the same upper limits on aperture size, regardless of format. For example, if one's limit is f/2.8 at 200mm, it is probably f/4 at 300mm and f/5.6 at 400mm. So if instead the smaller format or smaller pixels can be used with a lens of equally large aperture (diameter), leading to similar size, weight and cost, it will get

    1) aperture ratio lower in proportion to focal length,

and thus

    2) equal DOF (on equal sized prints)

and also

    3) lower ISO needed to get the same high shhutter speed,
        proportional to the area of the sensor and pixels

with the lower ISO setting in turn leading to

    4) equal amount of light gathered per pixel and thus equal "signal",
       measured in electron counts.

Adding the fact that

    5) smaller photosites produce if anything slightly LESS
       electrons of thermal noise, particularly in long exposures

we see that smaller pixels will then give about equal signal to noise ratio, and might be better in long exposures.


P. S. In a website with a leaning towards landscape photography, it puzzles me why so many people ignore or even actively deny how common it is to have to stop down for adequate DOF, at least when asserting the inferiority of smaller formats.
Logged

jani

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1624
    • Øyet
PIXEL DENSITY
« Reply #19 on: September 19, 2005, 05:28:01 pm »

Quote
I asked to confirm that you are interested in telephoto reach beyond 400mm with your 20D. (Me too: A TC and/or sensor with smaller pixel spacing is second on my wish list after a macro lens.)  That alone seems a good reason to hold out for your next DSLR to have pixel spacing no larger than the 6.4 microns of the 20D. So I presume you are waiting on an affordable Canon 24x36mm format DSLR with more than 20MP.
That would be the obvious upgrade step, if all we think of is the sensor properties.

I, for one, probably won't upgrade for less than a 100% gain in the total number of pixels, so that possible future 20+ MPx full 135-format sensor camera would be nice.

While it means I won't get a higher pixel density at the FOV I'm using now, it probably will change the way I use my current lenses.
Logged
Jan
Pages: [1]   Go Up