Lens Performance with New Generation DSLR

sanking · Dec 28, 2008

One of the issues people will have with the new generation DSLR cameras that pack a lot of pixels on the sensor is lens quality. This will be true of both 15 mp APS sensor cameras like the Canon 50D as well as full frame sensor cameras like the Sony A 900.

In resolution testing the Canon 5D with several different lenses I found that the limiting quality for resolution was the sensor, not the lens. The theoretical maximum in terms of resolution of the Canon 5D is about 60 lines per mm, and what you get in practice is about 52-55 lines per mm.

In testing the Canon 50D, which puts 4752 linear pixels on a 22.3 mm sensor, the limit to resolution is definitely the lens, not the sensor. Using several different zoom lenses, including a couple of the expensive L lenses, I only got a maximum of about 80 lines per millimeter, whereas with a prime 50mm f/1.8 I was able to get 90 lines per millimeter.

Just food for thought for anyone planning to invest in a 22-24 mp full sensor camera. You are going to need some good optics to take advantage of the sensor.

Sandy King

spotmeter · Mar 15, 2009

sanking said:
One of the issues people will have with the new generation DSLR cameras that pack a lot of pixels on the sensor is lens quality. This will be true of both 15 mp APS sensor cameras like the Canon 50D as well as full frame sensor cameras like the Sony A 900.

In resolution testing the Canon 5D with several different lenses I found that the limiting quality for resolution was the sensor, not the lens. The theoretical maximum in terms of resolution of the Canon 5D is about 60 lines per mm, and what you get in practice is about 52-55 lines per mm.

In testing the Canon 50D, which puts 4752 linear pixels on a 22.3 mm sensor, the limit to resolution is definitely the lens, not the sensor. Using several different zoom lenses, including a couple of the expensive L lenses, I only got a maximum of about 80 lines per millimeter, whereas with a prime 50mm f/1.8 I was able to get 90 lines per millimeter.

Just food for thought for anyone planning to invest in a 22-24 mp full sensor camera. You are going to need some good optics to take advantage of the sensor.

Sandy King

This is great news, as I have a lot of Zeiss and Canon primes!

pellicle · Mar 16, 2009

Sandy

sanking said:
One of the issues people will have with the new generation DSLR cameras that pack a lot of pixels on the sensor is lens quality.
...
Just food for thought for anyone planning to invest in a 22-24 mp full sensor camera. You are going to need some good optics to take advantage of the sensor.

very good points but I would raise the following

when using wide angle or photographing more distant subjects the subject level detail is going to be an issue. Testing is done (often enough) at a distance related to the focal length of the lens, when using wide angle (even 50mm) this can result in a need for much higher resolution than a telephoto (because feature detail will be smaller when photographing at normal distances). This is something which I cover in the latter part of this blog article. If chasing better wide angle performance I recommend photographers to begin looking at bigger formats (where you're using a longer lens too). I find that images take in with my 90mm on 4x5 hold more detail than images taken with my TS-E 24 on film (or even on a Digital). This point is not insignificant as the cost of better bodies (both outright and in requisite computing power to process images) is perhaps close to the costs of a reasonable 4x5 outfit.
the bar is always being raised, but need it be so? If I like the results I get from a lens then with a higher density sensor (5D to 5D Mkii) then I'm comfortable that I'll be still as happy with the image as it will not be degraded. Perhaps I'll be able to print it more with out software interpolation, but it won't become suddenly bad. With colour photography (leave black and white aside for now) I feel I'll get better colour images from a 5D than from many scans of 35mm film. So if I have a lens which was giving me good results from scans of film, then I'msure it'll produce images of sufficient quality (I've had good prints made to 40cm wide from 35mm) on digital. I feel that this equates to 5D or better simply being better and unless you're using a chumpy zoom (like say a cheapie 35-80) that good images which you liked from your lenses will remain good images that you like on a better body (should you upgrade from 5D to 5D MkII).

Granted if you find that your DSLR is showing up your chintzy lens to be just that (Tamron 28-200 anyone?) and you hadn't noticed it on film before then upgrading your camera may come as a surpirise (if so its been a good learning exersize), but I don't really think that my 12 year old EF200 f2.8 will suddenly suck when I put a 5D II behind it rather than a 5D or 20D. I could see issues on my poor lenses back when I tried a Canon EOS 350D and didn't see much difference on my 20D.

Example:

I bought an EOS 630 some (like 17) years ago, it came with an EF35-70 lens which I thought to be "ok" for a walk about lens. My preferred lenses were my EF50 1.8 and an EF24 f2.8 ... years later when I got a 20D I found that the EF35-70 was rather soft and lacked contrast (which I suspected but rarely noticed on negs) when compared to the EF 50.

To some extent this can be 'compensated for' with use at smaller apertures and by a little 50 pixel radius ~ 20% sharpening in photoshop. I don't think it will look worse (than it already does) when used on a 5D MkII than it does on a 5D just you'll have higher magnification to deal with.

Magnification will always reveal more flaws, but as I mentioned above if the prints were satisfying they probably still will be.

IMHO the greatest change will be seen in going from APS sensor to Full frame as this will bring the edges into the picture (and there my old EF35-70 really falls down).

sanking · Mar 20, 2009

Hi Pellicle and Spotmeter,

Thanks for your comments.

Since I posted the original message on this subject I spent several week in Oaxaca, Mexico working with the Canon 50D. I had with me several zooms and a couple of primes.

My working procedures are probably different from those of most in that I try to always use the camera on a tripod, and use aperture priority with the lens set at an aperture that is not diffraction limited. For the Canon 50D that is just over f/8. Of course, the primes will give much better performance at wider apertures.

What I found in practice is that at f/8 all of the lenses, zooms and primes alike, gave about the same performance. So, if you are able to use your camera on a tripod and set the aperture at about f/8 or lower you should be able to get excellent results with the Canon 50D with both primes and zooms.

About noise. Since I use the camera on a tripod whenever necessary, with ASA set to 100, noise with the 50D has not been a problem, as some reviews suggested it might be. However, I generally run my digital capture shots through a noise reducing program as I like a smooth, creamy type look in my prints. Heritage of LF I guess.

Sandy King

jd callow · Mar 21, 2009

My experience with noise has been that it increases with length of exposure and at the ends of the scale. I also use tripods and aperture priority.

pellicle · Mar 21, 2009

Sandy

sounds (to me here in Finland) like an exotic location trip, hope you had a good time

sanking said:
My working procedures are probably different from those of most in that I try to always use the camera on a tripod, and use aperture priority with the lens set at an aperture that is not diffraction limited.

sounds remarkably like mine ...

For the Canon 50D that is just over f/8. Of course, the primes will give much better performance at wider apertures.

ok .. interesting I have some general rule of thumb (though nothing clear) in my mind about this myself and thought it tended to get better by stopping down until you got to diffraction producing detraction of the image.

I think I'll do a few tests with my 50 f1.8 on my 10D (and publish it) to get a better 'feel for this' (thanks for the idea ;-)

What I found in practice is that at f/8 all of the lenses, zooms and primes alike, gave about the same performance. So, if you are able to use your camera on a tripod and set the aperture at about f/8 or lower you should be

ok ... again this matches my personal taste and operation with my camera, nice to see others find similar.

About noise. Since I use the camera on a tripod whenever necessary, with ASA set to 100, noise with the 50D has not been a problem, as some reviews suggested it might be. However, I generally run my digital capture shots through a noise reducing program as I like a smooth, creamy type look in my prints. Heritage of LF I guess.

interesting (the noise reduction bits), do you find that there is any significant loss of image quality? I was recently scanning some old film from my Japan years and found that it was interesting how once feature size shrank things turned from 'discernable detail' to fuzz. Specifically how broad leaf species and conifers differed.

anyway .. thanks for your test details

pellicle · Mar 21, 2009

JD

jd callow said:
My experience with noise has been that it increases with length of exposure and at the ends of the scale.

I seem to observe that my images when its been minus 10 C or less seem less noisy than when its "room temperatures" ... its a pitty that my batteries don't like that sort of temperature (and give up the ghost)

)

sanking · Mar 23, 2009

Well, with the Canon 50D because of the small pitch of the sensor points diffraction limiting aperture is at just over f/8. F/11 is ok, but f/16 gives considerable loss of sharpness.

BTW, this would not be the same for the 40D as the sensor points are larger.

Sandy

pellicle said:
ok .. interesting I have some general rule of thumb (though nothing clear) in my mind about this myself and thought it tended to get better by stopping down until you got to diffraction producing detraction of the image.

I

sanking · Mar 23, 2009

I think the noise reduction helps the image as it look a lot smoother if done right. A smooth image looks better to me than one with hard biting detail. Perhaps this is my large format background kicking in!

Sandy King

pellicle said:
interesting (the noise reduction bits), do you find that there is any significant loss of image quality? I was recently scanning some old film from my Japan years and found that it was interesting how once feature size shrank things turned from 'discernable detail' to fuzz. Specifically how broad leaf species and conifers differed.

anyway .. thanks for your test details

pellicle · Mar 24, 2009

Hi

sanking said:
I think the noise reduction helps the image as it look a lot smoother if done right. A smooth image looks better to me than one with hard biting detail.

on a tangent here thinking about the advantages of sensor density to record more of what the lens can produce has me thinking of the Panasonic G1. I suspected that it would do well, and looking between images on DPReview it indeed seems that the 100iso image clarity of that will go close to the 5D.

Some years of using tiny compact cameras for snapshots (and metering for my 4x5) has spoiled me into wanting a compact which is better than 35mm can be and not much bigger to carry than my Pentax digital spotmeter (which now gets left at home most of the time)

I think I'm going to be switching from my Canon outfit soon

Loris Medici · Mar 24, 2009

Hi Sandy,

How do you calculate this? I use the following formula:

ADR = 1.22 x L x F

ADR: Airy Disk - Circle of Confusion - diameter (in mm)
L: Wavelength of light (in mm)
F: F stop

Do you use a airy disk diameter equal to the pixel pitch? For digital sensors with mosaic pattern using a ADR diameter of 2x pixel pitch would be more realistic; because only then you can be sure that you're registering full "color" detail.

Pixel pitch for 50D should be something like 0.0047mm (not sure) ->
0.0047 x 2mm = 1.22 x 0.000550mm x F ->
F = 14 (for green light - 0.000550mm)

Theorically, you should be able to fetch pretty good detail up to F11 (to be conservative) with 50D???

Do you observe any quality decrease - significant in real-life situations - between F8 and F11?

Regards,
Loris.

sanking said:
Well, with the Canon 50D because of the small pitch of the sensor points diffraction limiting aperture is at just over f/8. F/11 is ok, but f/16 gives considerable loss of sharpness.

BTW, this would not be the same for the 40D as the sensor points are larger.

Sandy

amphoto · Mar 24, 2009

Do sensors out-resolve lenses

Link to the 'Do sensors out-resolve lenses?' article that appeared on the Luminous-Landscape a while back:

http://www.luminous-landscape.com/tutorials/resolution.shtml

Warning: It's very technical. If you have a PhD in optical physics you should be ok - for the rest of us, well, at least the conclusion is short!

pellicle · Mar 24, 2009

Hi

amphoto said:
Link to the 'Do sensors out-resolve lenses?' article that appeared on the Luminous-Landscape a while back:

http://www.luminous-landscape.com/tutorials/resolution.shtml

there are some interesting points raised and some assumptions made which can't be ignored (like do you want your blue channel blury?) I thought that the fence sitting was done well in the conclusion Eg:

first this

Consider a 35mm system with a lens at f/11. At best, the maximum resolution you will get is equivalent to 16 MP, even if your camera has 22 or 25 MP. In the case of an APS-C based system the limit goes to 7 MP, and 4 MP considering a Four Thirds format. Stopping down to f/22 the limit of the effective resolution of the 35mm based system goes to 4 MP!

and then this

See again the Figure 2: the lens limits the resolution of the 5 microns pixel based system with an aperture of f/22, but it is also the case for f/16, f/11 or even f/8. That pixel pitch leads to a 10 MP Four Thirds sensor, a 15 APS-C sensor, a 35 MP sensor of 35mm format and a 70 MP sensor of 36x48mm dimension.

the 10MP four thirds and 15MP APS is interesting as that's what the G1 and 50D are ...

Personally (if the DPReview tests are not fudged) I do see increases in resolution in the studio shots (using a 50mm lens) between 10D - 20D - 40D and 50D.

Do I want to pay US$3500 for a 5D MkII (body only) or will I be content with a G1 for US$650 (nearly five times less and comes with a useful lens).

Loris Medici · Mar 24, 2009

pellicle said:
...

the 10MP four thirds and 15MP APS is interesting as that's what the G1 and 50D are ...
...

Do I want to pay US$3500 for a 5D MkII (body only) or will I be content with a G1 for US$650 (nearly five times less and comes with a useful lens).

Pellicle, I don't get your point? Given you have a perfect lens on both cameras, with 50D, you'll have an extra 5 megapixels (50% more!) to print from (with equal pixel to pixel sharpness), when compared to G1 (which I must admit I don't know anything about...)

The question then transforms to: "Do I need the extra megapixels?", which transforms to: "Do I need to print bigger?" and/or "Do I need the luxury cropping - more - freely in the post-processing phase?"

Unfortunately, after a level, marginal increase in convenience comes only coupled with serious expenses... The real / final question is: "Do I really need that extra convenience... badly?"

sanking · Mar 24, 2009

Loris,

I have not done any calculations myself but various sources show that the diffraction limited aperture of the 50D is about f/8, including this one.

http://photocamel.com/forum/canon-f...-diffraction-limited-aperture-calculated.html

There is a link in there to how the calculations were done.

From a more practical perspective, I tested the camera with a resolution target using several different lenses. There was very little difference in resolving power between results at f/8 and at f/11, but significant loss at f/16.

Sandy

donbga · Mar 24, 2009

sanking said:
Loris,

I have not done any calculations myself but various sources show that the diffraction limited aperture of the 50D is about f/8, including this one.

http://photocamel.com/forum/canon-f...-diffraction-limited-aperture-calculated.html

There is a link in there to how the calculations were done.

From a more practical perspective, I tested the camera with a resolution target using several different lenses. There was very little difference in resolving power between results at f/8 and at f/11, but significant loss at f/16.

Sandy

I think this thread found on the Fred Miranda forums is sort of interesting to read:

http://www.fredmiranda.com/forum/topic/695026/0

Don Bryant

tom_micklin · Mar 25, 2009

pellicle said:
Hi

Do I want to pay US$3500 for a 5D MkII (body only) or will I be content with a G1 for US$650 (nearly five times less and comes with a useful lens).

Hi,
I'm wondering where you are seeing the price of the 5D at US$3500. Both of my local camera stores (Seattle) - and BH photo online list it at $2699.[/U]

pellicle · Mar 25, 2009

Tom

tom_micklin said:
Hi,
I'm wondering where you are seeing the price of the 5D at US$3500. Both of my local camera stores (Seattle) - and BH photo online list it at $2699.[/U]

look again, you'll see Mk II at the end of the 5D in my post ... BH online lists that at US$3499 when I wrote it.

pellicle · Mar 25, 2009

Hi

Loris Medici said:
Pellicle, I don't get your point?

well there was a couple, my first was the apparent fence sitting of the article which I quoted from ... saying on one hand this then on the other hand that.

then my other point was something like this:

I think that a G1 Panasonic will produce images of comparable quality to a 5D, since that camera was (and still is) a reasonable bench mark of quality then perhaps a G1 will do me, and I do not then need to pop 5 times that price on a 5D Mk II just to better it by some margin (especially when I have 4x5 gear already).

To clarify, the G1 (due to size) makes a better "carry it all the time" camera to take the best image with available equipment (given that I'd leave my 3kg of DSLR gear at home some times) and that if I am planning a photograph then my 4x5 system may just give me a result which is comparable to that obtained with a 5D MkII

Given you have a perfect lens on both cameras, with 50D, you'll have an extra 5 megapixels (50% more!) to print from (with equal pixel to pixel sharpness), when compared to G1 (which I must admit I don't know anything about...)

well I'd argue that 5 megapixels is not 50% more mainly because its got to be thought of as a square; 3x4 = 12 and 4x5 = 20 but its only an extra pixel even though it seems to be 60% more. So, given that

the 50D yields an image which is 4752 x 3168 and
that the G1 yields one which which is 4000 x 3000

so then printing an image to 300dpi (to pick a figure) would give a print of:

15.8 x 10.56 inches for the 5D and
13.3 x 10 inches for the G1

so if that's 50% bigger to you then we have a different understanding of percent

If you don't know it, the G1 is a compact 4/3'rds camera (meaning that the sensor is the same as in the Olympus range of DSLRs) which seems to answer my prayers for combining compact body + acceptably large sensor (more details here and my own head scratching about why I want it here, here and here.)

The question then transforms to: "Do I need the extra megapixels?", which transforms to: "Do I need to print bigger?" and/or "Do I need the luxury cropping - more - freely in the post-processing phase?"

Unfortunately, after a level, marginal increase in convenience comes only coupled with serious expenses... The real / final question is: "Do I really need that extra convenience... badly?"

well now its my turn to ask ... I don't quite get your point here ;-)

pellicle · Mar 25, 2009

Sandy

sanking said:
From a more practical perspective, I tested the camera with a resolution target using several different lenses. There was very little difference in resolving power between results at f/8 and at f/11, but significant loss at f/16.

interesting ... I'll have a peek myself

on another note, I thought I'd laugh myself silly and test my Coolpix 5000 (which is a 5MP digital camera). It has a 7 ~ 21 mm lens which on its 2/3" sensor gives about 28mm effective for 35mm film (gosh I wish we could use angle of view ... instead of all these messy mm bits).

Due to the requirements of lens testing (placing the lens a distance related to its focal length away from the chart) I get whacko results like well over 100lp/mm and nearly into the 200 lp/mm, but when I wack it on a wall that's 5 meters away (about the normal distance for normal indoor wide angle shots and also coincidentally about right for testing my Bessa RF with its 105mm lens) then I get about 5lp/mm in the picture (while the Bessa gets somewhere around 20lp/mm).

Funny stuff lens testing I still reckon we need something more understandable for testing standards (so casual readers can comprehend the specs and relate it to their own pictures).

Loris Medici · Mar 25, 2009

pellicle said:
...

well I'd argue that 5 megapixels is not 50% more mainly because its got to be thought of as a square; 3x4 = 12 and 4x5 = 20 but its only an extra pixel even though it seems to be 60% more. So, given that

the 50D yields an image which is 4752 x 3168 and

that the G1 yields one which which is 4000 x 3000

so then printing an image to 300dpi (to pick a figure) would give a print of:

15.8 x 10.56 inches for the 5D and

13.3 x 10 inches for the G1

so if that's 50% bigger to you then we have a different understanding of percent

Well, I assumed G1 was 10Mp, as that's exactly what one would understand from your previous message. If I knew about G1 (that it was a 12Mp camera!) then I wouldn't have said something like that -> victim of misinformation...

pellicle said:
well now its my turn to ask ... I don't quite get your point here ;-)

It's simple -> stop whining, buy a G1 and be happy instead.

Regards,
Loris.

pellicle · Mar 25, 2009

Loris Medici said:
It's simple -> stop whining, buy a G1 and be happy instead.

none in the shops here right now, and I'm just wondering if I'll sell my 10D or (since its worth not much) just keep it (what for??)

but yep, thats my direction

tom_micklin · Mar 25, 2009

pellicle said:
Tom

look again, you'll see Mk II at the end of the 5D in my post ... BH online lists that at US$3499 when I wrote it.

I did see that, but BH quotes that price for the camera plus lens. You said Body Only, which was why I wondered.
Tom

pellicle · Mar 25, 2009

Ok

tom_micklin said:
I did see that, but BH quotes that price for the camera plus lens. You said Body Only, which was why I wondered.
Tom

so I looked again and found that I looked at the wrong line ... its indeed US$2,699 body only. I looked quickly and that seemed about the right number 9and I keep needing to check prices in other places as I do not live in USA. I'm Australian and in my money they're nearly $4000 body only (which is still a bit dearer than BH charges). Worse, here in Finland (where I am now) they cost 3200 with the kit lens and over 2700 Euro body only.

but thanks for pointing it out ... I'm now not as scared by the price

(I think I'll still get a Panasonic G1 and retire my 10D though)

sanking · Mar 25, 2009

You have to take into account sensor size and the fact that for the same depth of field you are using optimum apertures. I have not tested it but I am fairly confident that my Canon G9 can resolve somewhere around 200 l/mm in the center of the sensor.

But to make a print 12X18 in size you have to blow up that 7mm wide sensor image a lot.

Sandy

pellicle said:
Sandy

interesting ... I'll have a peek myself

on another note, I thought I'd laugh myself silly and test my Coolpix 5000 (which is a 5MP digital camera). It has a 7 ~ 21 mm lens which on its 2/3" sensor gives about 28mm effective for 35mm film (gosh I wish we could use angle of view ... instead of all these messy mm bits).

Due to the requirements of lens testing (placing the lens a distance related to its focal length away from the chart) I get whacko results like well over 100lp/mm and nearly into the 200 lp/mm, but when I wack it on a wall that's 5 meters away (about the normal distance for normal indoor wide angle shots and also coincidentally about right for testing my Bessa RF with its 105mm lens) then I get about 5lp/mm in the picture (while the Bessa gets somewhere around 20lp/mm).

Funny stuff lens testing I still reckon we need something more understandable for testing standards (so casual readers can comprehend the specs and relate it to their own pictures).

Lens Performance with New Generation DSLR

Moderator

Privacy & Transparency

Privacy & Transparency