Luminous Landscape Forum

Equipment & Techniques => Cameras, Lenses and Shooting gear => Topic started by: fike on December 11, 2008, 12:30:18 PM

Title: Interesting Comparison
Post by: fike on December 11, 2008, 12:30:18 PM
Over the last few months we have been discussing the relative merits of the new high-Mega Pixel cameras.  the 50D gained the dubious notariety of the camera that went "a bridge too far," just too many megapixels for its sensor size.  

I shoot a 30D now, and I've had the 50D in my sights for a few months.  I have been waiting for the stars to align and the price to decline.  But I have constantly had in the back of my mind the issue of more megapixels versus good megapixels.

Now comes the 5dMkII.  It really is a bit out of my price-range.  But that problem of good megapixels keeps bedeviling me.  It occurred to me yesterday that the 5dMkII has the same pixel pitch as the 20D/30D cameras.  

--so here's the controversial statement--

What this means to me is that when I put my 100-400 f/4.5-5.6 L IS lens on a 5dMkII, under ideal conditions and at base ISOs, I will get very similar resolution out of the same cropped field of view subject between the 30D and the 5dMkII cameras.

Yes, color, noise at higher ISOs, dynamic range, and a myriad of other factors will be better in the 5dMkII, but for sheer reach out of a telephoto zoom lens and camera combination, the 5dMkII and 30D will be equivalent. The 50D wins in this narrow category.  What I am really curious to see is controlled comparisons of the previous generation 20/30D and the 5dMkII.  The 5dMkII will obviously be better, but how much?  Will resolution numbers be the same when factoring in the crop-factor?

To get a similar telephoto reach and image quality out of a 5dMkII, you would need to invest many thousands more in a high-quality longer lens.  This makes the price differential even higher between the 50D solution and the 5dMkII solution.

What are your thoughts on my logic?
Title: Interesting Comparison
Post by: mike.online on December 11, 2008, 05:00:42 PM
I also hasve a 30D and would like to see a comparison. as for your longer lens question... you could just use a 1.4 or 2.0 extender. the 1.4 will get you pretty close to the APS-C magnification factor on a full frame sensor.
Title: Interesting Comparison
Post by: fike on December 11, 2008, 05:13:25 PM
Quote from: mike.online
I also hasve a 30D and would like to see a comparison. as for your longer lens question... you could just use a 1.4 or 2.0 extender. the 1.4 will get you pretty close to the APS-C magnification factor on a full frame sensor.

The 1.4x teleconverter is a decent idea, but again, that is a decrease in image quality and a loss of a single stop of light.  Will this cancel out the benefit of the improved body/sensor? Would there be unacceptable increased vignettingand distortion more obviously displayed by the full frame sensor and further exacerbated by the teleconverter.  

So an interesting hypothesis to test would be:

The 5dMkII with a 1.4x teleconverter has equivalent resolution and quality to a 20/30D without teleconverter???
Title: Interesting Comparison
Post by: Panopeeper on December 11, 2008, 05:25:12 PM
Quote from: fike
To get a similar telephoto reach and image quality out of a 5dMkII, you would need to invest many thousands more in a high-quality longer lens

I don't understand this. You explained just above, that if you crop away the edges of a 5D2 image, you get the same resolution and same angle of view as with the 20D/30D, using the very same lens. The pixel quality of the 5D2 is definitively better than that of the 20D-50D.

So, why would you need a different lens?
Title: Interesting Comparison
Post by: fike on December 11, 2008, 05:33:29 PM
Quote from: Panopeeper
I don't understand this. You explained just above, that if you crop away the edges of a 5D2 image, you get the same resolution and same angle of view as with the 20D/30D, using the very same lens. The pixel quality of the 5D2 is definitively better than that of the 20D-50D.

So, why would you need a different lens?

To be more precise, I should have said:
Quote
To get a similar telephoto reach and the full benefit of the improved image quality(and resolution) out of a full-frame 5dMkII, you would need to invest many thousands more in a similarly high-quality longer lens to get the same effective 640mm range at full-frame coverage.

As you probably know, the price differential between a decent 600mm lens and 400mm lens is quite substantial.

This analysis really is predicated on the importance of long telephoto reach.  The opposite argument could certainly be made regarding wide-angle capabilities.  I am obviously weighting long telephoto as more important.
Title: Interesting Comparison
Post by: Panopeeper on December 11, 2008, 06:33:08 PM
Quote from: fike
To get a similar telephoto reach and the full benefit of the improved image quality(and resolution) out of a full-frame 5dMkII, you would need to invest many thousands more in a similarly high-quality longer lens to get the same effective 640mm range at full-frame coverage.

Ok, what do you need the more pixels for?

1. To make larger prints? Is the higher pixel level quality of the 5D2 not enough to print with lower resolution (a kind of upresing)?

2. To downres the image in order to increase the quality? Would that be necessary even with the 5D2 images?

What I am trying to arrive at is, that a crop from the 5D2 image is more worth than a full 20D-30D-40D image, perhaps even than a full 50D image. Keep in mind, that downresing the 50D images reduces not the noise but its visibility, and it does not increase the dynamic range.
Title: Interesting Comparison
Post by: fike on December 11, 2008, 06:44:54 PM
Quote from: Panopeeper
Ok, what do you need the more pixels for?

1. To make larger prints? Is the higher pixel level quality of the 5D2 not enough to print with lower resolution (a kind of upresing)?

2. To downres the image in order to increase the quality? Would that be necessary even with the 5D2 images?

What I am trying to arrive at is, that a crop from the 5D2 image is more worth than a full 20D-30D-40D image, perhaps even than a full 50D image. Keep in mind, that downresing the 50D images reduces not the noise but its visibility, and it does not increase the dynamic range.

You have framed the question perfectly.  I would be comfortable saying that a 5dMkII crop would be better than the equivalent field of view from a 20D/30D--similar number of pixels, better quality from improved contrast, color, decreased SNR.  

While I am comfortable saying that the 5dMkII will have better quality, I am interested to see how that difference manifests itself.  

The 40d/50d in that same scenario makes an interesting discussion. I would find it informative to see the same cropped field of view from all the cameras adjacent to one another.  

You have got me pegged. I like to print very large images--24"x56".  I generally shoot panoramic or mosaic, so wide angle capability is not a problem. I can always stitch the image.
Title: Interesting Comparison
Post by: fike on December 11, 2008, 08:35:30 PM
Very interesting comparison data at the Digital Picture:

The Digital Picture 5DMKII Review (http://www.the-digital-picture.com/Reviews/Canon-EOS-5D-Mark-II-DSLR-Digital-Camera-Review.aspx)

I find his comparative analysis of DLA--Diffraction Limited Aperture to be very interesting.
Title: Interesting Comparison
Post by: Ray on December 12, 2008, 12:12:02 AM
Quote from: Panopeeper
I don't understand this. You explained just above, that if you crop away the edges of a 5D2 image, you get the same resolution and same angle of view as with the 20D/30D, using the very same lens. The pixel quality of the 5D2 is definitively better than that of the 20D-50D.

So, why would you need a different lens?

How much better is the pixel quality, Gabor? I'd like to see this. In fact, such a comparison would be very revealing of just how much improvement has taken place at the pixel level, during the past few years.

The 40D has very liitle improvement over the 20D, when comparing equal size images, and the 50D has perhaps marginally more improvement over the 40D, when comparing equal image sizes.

The pixel size of the 5D2 is roughly equal to the pixel size of the 20D. Comparing equal size crops comprised of the same number of pixels, a 5D2 image will be better, no doubt. But how much better is the 5D2 pixel? This appears to be an unknown quantity. Can you post some comparisons?
Title: Interesting Comparison
Post by: Ray on December 12, 2008, 09:18:23 AM
DXOmark has some interesting information comparing the 20D with the 5D2, which I understand is at the pixel level. According to their graphs, SNR of the 5D2 at ISO 1600 is just 1.4db better than the 20D. That doesn't sound much, especially when you consider that ISO 1600 on the 5D2 is actually ISO 1093 and on the 20D actually ISO 1333. In other words, I take a fully exposed shot with the 20D at ISO 1600, then take the same shot with the same lens on the 5D2, using the same exposure at ISO 1600, crop the 5D2 image to the same FoV as the 20D shot, and that 1.4db S/N advantage disappears because the 5D2 shot is slightly underexposed.
Title: Interesting Comparison
Post by: fike on December 12, 2008, 10:02:05 AM
Quote from: Ray
DXOmark has some interesting information comparing the 20D with the 5D2, which I understand is at the pixel level. According to their graphs, SNR of the 5D2 at ISO 1600 is just 1.4db better than the 20D. That doesn't sound much, especially when you consider that ISO 1600 on the 5D2 is actually ISO 1093 and on the 20D actually ISO 1333. In other words, I take a fully exposed shot with the 20D at ISO 1600, then take the same shot with the same lens on the 5D2, using the same exposure at ISO 1600, crop the 5D2 image to the same FoV as the 20D shot, and that 1.4db S/N advantage disappears because the 5D2 shot is slightly underexposed.


That's pretty interesting.  Now what I want to see is how this minor difference manifests itself in images.  Most of the comparisons between full-frame and APS-C have focused on comparing the full-frame to the cropped frame.  What is interesting about the identical pixel pitch and similar SNR is that we can now look at a level playing field by cropping the full-frame sensor down to the APS-C sized field of view.  

I think this will shed light on the issues comparing the 50D and the 40D cameras.

But again, this allows us to view comparison images and see the effect of "better pixels" without confusing the issue with megapixel and pixel-pitch discussions.
Title: Interesting Comparison
Post by: Ray on December 12, 2008, 11:12:21 AM
Quote from: fike
That's pretty interesting.  Now what I want to see is how this minor difference manifests itself in images.  Most of the comparisons between full-frame and APS-C have focused on comparing the full-frame to the cropped frame.  What is interesting about the identical pixel pitch and similar SNR is that we can now look at a level playing field by cropping the full-frame sensor down to the APS-C sized field of view.  

I think this will shed light on the issues comparing the 50D and the 40D cameras.

But again, this allows us to view comparison images and see the effect of "better pixels" without confusing the issue with megapixel and pixel-pitch discussions.

The graphs at DXOmark overlay an image which appears if you move the cursor to the colored bar on the right. It's easy to see how the noise increases, with all cameras compared, as you change ISO, ie. by moving the cursor up and down the colored bar.

I'd estimate the 5D2 has just under a stop of S/N improvement over the 50D at most ISOs. However, a 5D2 image cropped to the FoV of a 50D image becomes only 8mp. As mentioned before, most of that 3/4 stop advantage will disappear if the 50D images is downsampled to 8mp, or, if the 8mp crop is interpolated to 15mp. The net effect is either better resolution from the 50D or equal noise, approximately.

http://www.dxomark.com/index.php/eng/Image.../(brand3)/Canon (http://www.dxomark.com/index.php/eng/Image-Quality-Database/Compare-cameras/(appareil1)/279%7C0/(appareil2)/180%7C0/(appareil3)/267%7C0/(onglet)/0/(brand)/Canon/(brand2)/Canon/(brand3)/Canon)
Title: Interesting Comparison
Post by: fike on December 12, 2008, 11:36:06 AM
Quote from: Ray
The graphs at DXOmark overlay an image which appears if you move the cursor to the colored bar on the right. It's easy to see how the noise increases, with all cameras compared, as you change ISO, ie. by moving the cursor up and down the colored bar.

I'd estimate the 5D2 has just under a stop of S/N improvement over the 50D at most ISOs. However, a 5D2 image cropped to the FoV of a 50D image becomes only 8mp. As mentioned before, most of that 3/4 stop advantage will disappear if the 50D images is downsampled to 8mp, or, if the 8mp crop is interpolated to 15mp. The net effect is either better resolution from the 50D or equal noise, approximately.

http://www.dxomark.com/index.php/eng/Image.../(brand3)/Canon (http://www.dxomark.com/index.php/eng/Image-Quality-Database/Compare-cameras/(appareil1)/279%7C0/(appareil2)/180%7C0/(appareil3)/267%7C0/(onglet)/0/(brand)/Canon/(brand2)/Canon/(brand3)/Canon)


I saw that overlay. I assumed that it was merely for representation only...not real data-based images.  

Your conclusion about the 50D having better resolution within the cropped sensor field of view and nearly equal noise is likely to be debated heavily, although I think you are probably right.  As I keep repeating, I want to see what the difference really looks like in the real world.  Everyone wants to uprez or downrez the images to compare.  In this comparison, no resizing is needed.  since a 20/30d share the same pixel pitch cropping is sufficient.  When the 50D is thrown into the mix, then you need to do some sort of resampling, either up for the 5dMkII or down for the 50D.  

Another aspect to this debate is when you start throwing High-ISO into the mix.  How high do you need to go before the 5dMkII overtakes the 50D in cropped sensor area resolution?
Title: Interesting Comparison
Post by: Panopeeper on December 12, 2008, 02:15:55 PM
Quote from: Ray
How much better is the pixel quality, Gabor? I'd like to see this. In fact, such a comparison would be very revealing of just how much improvement has taken place at the pixel level, during the past few years

I did not qualify my statement (I was blathering).

Those, who need the maximum DR, like landscapers, shoot mostly at the native ISO or in the low range. The DR of the 5D2 seems to be only negligably greater than that of the 40D @ ISO 100.

For those, who are after available-light performance, higher ISO is important. The 5D2 is about one stop better than the 40D @ ISO 1600.

I ordered a Stouffer transmission wedge; as soon as it arrives, I will make measurements of the 20D and the 40D, and I hope to find some photogs in my area, who would come by with their cameras and make suitable shots, so that I can build a large camera base. I am disappointed with the raw files posted at Imaging Resource; they are obviously not using those images for accurate measurements.

I don't believe the more accurate measures will differ very much from those I made until now, but those results are much more reliable. I have such raw files for the 5D (but only ISO 100), the Nikon D200 (only ISO 100) and the D3 (full serie). Those shots prove, that the measurement is very consistent within a small margin.
Title: Interesting Comparison
Post by: Ray on December 12, 2008, 10:21:25 PM
Quote from: fike
As I keep repeating, I want to see what the difference really looks like in the real world.

Prints or Plasma/LCD display?

At present, I'm converting a good few of my images to sRGB and 1920x1080 pixels for display on my new 50" Plasma TV. Where I can crop to the 16:9 aspect ratio without spoiling the composition, I do so.

It so happens I have a mounted print hanging on the wall immediately above the plasma screen, to one side. The scene is a wide-angle shot of a temple ruin at Angkor Wat taken with my 5D and printed on my Epson 7600 at 23" x35", no cropping, the full 12.7mp interpolated.

Out of curiosity, I took the same image file that I'd printed at 23"x35", converted it to sRGB and downsampled it to 1080 pixels in height, which resulted in a 5mb file (not 5mp but 5MB which is less than 2mp).

From a 'normal' (or shall we say 'usual') viewing distance of 10 or 12ft, the 5MB file on the Plasma screen looks just as detailed as the 260MB file (after interpolation) that was printed.

I'm a bit flabbergasted. Is there something wrong with my eyesight? I need glasses for close reading, but long distance is no problem. However, medium distances can be a slight problem. I can drive a car at night with no problem, without glasses, but for best clarity on the computer monitor I need galsses with a lower magnification than my normal reading glasses. For best clarity at medium distances, I use the very first spectacles I was prescribed for reading purposes, with a magnification of 1x.

I have no reason to think that this effect I'm seeing is a result of poor eyesight, but it might be. Maybe I'm in a state of delusion.

The other impression one clearly gets, making this comparison between the print of the 260MB file and the 5MB plasma display of the same file, is that the vibrancy and luminosity of the image on the plasma screen completley trounces the relatively dull and flat print. I'm beginning to think that the only purpose of my Epson 7600 printer is to produce very wide panoramas of stitched images that my plasma display cannot accommodate without very severe reduction in resolution and size.

However, I don't want to appear to be exaggerating. The appearance of the print changes with lighting. Sometimes in the evening, with a certain type of artificail lighting, the qualities change and subtleties of tonality are quite enchanting on the print. The plasma display is 'full on', as it were, day or night.
Title: Interesting Comparison
Post by: Nacnud on December 15, 2008, 06:53:52 AM
I had a conversation yesterday that kinda gels with with thread.
My new LX3 has an interesting feature, when you select a smaller image size (e.g. 6Mp instead of 10Mp) you get an increased zoom range!
I does this by zooming electronically until the sensor crop shrinks to the desired number of pixels, then it zooms optically.
As well as giving me an extended zoom range it means the sensor is avoiding any soft edge effect from the lens.
Very cunning!

Now imagine if the 5D2 on sRAW gave you the option of either a resized full frame or cropped frame to get the smaller file.
This would give you the option of sacrificing pixels in exchange for a much greater zoom range.

I realise I can do this by cropping in post-processing, but in terms of a composition aid and checking focus it would be great if the cropped image filled the frame.
Maybe the LX3 has something that could useful be applied to DSLRs too!
Title: Interesting Comparison
Post by: Ray on December 15, 2008, 08:25:52 AM
Quote from: Nacnud
I had a conversation yesterday that kinda gels with with thread.
My new LX3 has an interesting feature, when you select a smaller image size (e.g. 6Mp instead of 10Mp) you get an increased zoom range!
I does this by zooming electronically until the sensor crop shrinks to the desired number of pixels, then it zooms optically.
As well as giving me an extended zoom range it means the sensor is avoiding any soft edge effect from the lens.
Very cunning!

Now imagine if the 5D2 on sRAW gave you the option of either a resized full frame or cropped frame to get the smaller file.
This would give you the option of sacrificing pixels in exchange for a much greater zoom range.

I realise I can do this by cropping in post-processing, but in terms of a composition aid and checking focus it would be great if the cropped image filled the frame.
Maybe the LX3 has something that could useful be applied to DSLRs too!

There is actually an advantage in having that wider view when taking the shot. I find often when using a 400mm lens with a cropped format camera like the 20D it's difficult to locate the target, especially when the tareget is a small bird flitting from branch to branch.
Title: Interesting Comparison
Post by: ErikKaffehr on December 15, 2008, 09:06:55 AM
Hi,

I think this is mostly a question of viewing distance. The resolution of the human eye is regarded to be normally about one minute of arc, and be determined by the average distance between the rods on the foeva. If my calculations are right the resolution of the plasma TV would match the resolution of the eye at about 2m (or 6 feet). So if you look at your print at a distance which is shorter than two neters it should be sharper than your plasma TV.

Another issue is that prints need a lot of light, 400 lux or so. Prints also have better permanence than plasma screens :-)

Best regards
Erik


Quote from: Ray
Prints or Plasma/LCD display?

At present, I'm converting a good few of my images to sRGB and 1920x1080 pixels for display on my new 50" Plasma TV. Where I can crop to the 16:9 aspect ratio without spoiling the composition, I do so.

It so happens I have a mounted print hanging on the wall immediately above the plasma screen, to one side. The scene is a wide-angle shot of a temple ruin at Angkor Wat taken with my 5D and printed on my Epson 7600 at 23" x35", no cropping, the full 12.7mp interpolated.

Out of curiosity, I took the same image file that I'd printed at 23"x35", converted it to sRGB and downsampled it to 1080 pixels in height, which resulted in a 5mb file (not 5mp but 5MB which is less than 2mp).

From a 'normal' (or shall we say 'usual') viewing distance of 10 or 12ft, the 5MB file on the Plasma screen looks just as detailed as the 260MB file (after interpolation) that was printed.

I'm a bit flabbergasted. Is there something wrong with my eyesight? I need glasses for close reading, but long distance is no problem. However, medium distances can be a slight problem. I can drive a car at night with no problem, without glasses, but for best clarity on the computer monitor I need galsses with a lower magnification than my normal reading glasses. For best clarity at medium distances, I use the very first spectacles I was prescribed for reading purposes, with a magnification of 1x.

I have no reason to think that this effect I'm seeing is a result of poor eyesight, but it might be. Maybe I'm in a state of delusion.

The other impression one clearly gets, making this comparison between the print of the 260MB file and the 5MB plasma display of the same file, is that the vibrancy and luminosity of the image on the plasma screen completley trounces the relatively dull and flat print. I'm beginning to think that the only purpose of my Epson 7600 printer is to produce very wide panoramas of stitched images that my plasma display cannot accommodate without very severe reduction in resolution and size.

However, I don't want to appear to be exaggerating. The appearance of the print changes with lighting. Sometimes in the evening, with a certain type of artificail lighting, the qualities change and subtleties of tonality are quite enchanting on the print. The plasma display is 'full on', as it were, day or night.
Title: Interesting Comparison
Post by: Ray on December 15, 2008, 10:28:21 AM
Quote from: ErikKaffehr
Hi,

I think this is mostly a question of viewing distance. The resolution of the human eye is regarded to be normally about one minute of arc, and be determined by the average distance between the rods on the foeva. If my calculations are right the resolution of the plasma TV would match the resolution of the eye at about 2m (or 6 feet). So if you look at your print at a distance which is shorter than two neters it should be sharper than your plasma TV.

Another issue is that prints need a lot of light, 400 lux or so. Prints also have better permanence than plasma screens :-)

Best regards
Erik

Erik,
Yes, it's definitely a matter of viewing distance. Your calculation of about 2 metre's viewing distance seems about right, but in practice I'm usually further away from the print than that, most of the time.

I guess a print would have better permanence than a plasma display which was on all the time. However, Panasonic are boasting that their latest displays will last longer than most of their customers who buy them. They are rated at 100,000 hours which, at 8 hours a day, works out to 34 years. At a more reasonable 4 hours a day, that would be 68 years   .
Title: Interesting Comparison
Post by: ErikKaffehr on December 15, 2008, 12:08:40 PM
;-) (Re. permanence)

Quote from: Ray
Erik,
Yes, it's definitely a matter of viewing distance. Your calculation of about 2 metre's viewing distance seems about right, but in practice I'm usually further away from the print than that, most of the time.

I guess a print would have better permanence than a plasma display which was on all the time. However, Panasonic are boasting that their latest displays will last longer than most of their customers who buy them. They are rated at 100,000 hours which, at 8 hours a day, works out to 34 years. At a more reasonable 4 hours a day, that would be 68 years   .
Title: Interesting Comparison
Post by: fike on December 15, 2008, 01:29:42 PM
Quote from: ErikKaffehr
;-) (Re. permanence)

I like to print big, and I like to print detailed.  I typically think of a scenario where the person views the image from a meter or two and then walks in close to inspect it more closely...perhaps from 6 inches.

LCDs and Plasmas are beautiful but don't hold up to this kind of close inspection.  I have a mosaic stitched print that is made in three 24" wide strips. The overall print is 6 feet wide by 4 feet tall.  You can see people in the image that are less than 3mm high.

As for permanence, I am not sold on the temporary LCD/Plasma displays.  I guess I am just old fashioned.  

so, for long-lens detail capture, I think the 50D may be superior to the 5DMkII under ideal conditions or on a tripod.
Title: Interesting Comparison
Post by: ErikKaffehr on December 15, 2008, 03:35:45 PM
Good observation, Ray!

On the whole the 5DII is probably better than the 20D, I'd guess, due to weaker antialiasing filter. Have you seen this page:

http://www.wlcastleman.com/equip/reviews/40D/index.htm (http://www.wlcastleman.com/equip/reviews/40D/index.htm) ?

BTW, my Sony SAL 24-70/2.8 ZA is on the mail to my great surprise so I guess I can post some findings after the weekend on that lens on the Sony Alpha 900.

Best regards
Erik


Quote from: Ray
DXOmark has some interesting information comparing the 20D with the 5D2, which I understand is at the pixel level. According to their graphs, SNR of the 5D2 at ISO 1600 is just 1.4db better than the 20D. That doesn't sound much, especially when you consider that ISO 1600 on the 5D2 is actually ISO 1093 and on the 20D actually ISO 1333. In other words, I take a fully exposed shot with the 20D at ISO 1600, then take the same shot with the same lens on the 5D2, using the same exposure at ISO 1600, crop the 5D2 image to the same FoV as the 20D shot, and that 1.4db S/N advantage disappears because the 5D2 shot is slightly underexposed.
Title: Interesting Comparison
Post by: John Chardine on December 15, 2008, 08:15:13 PM
Quote from: fike
The 1.4x teleconverter is a decent idea, but again, that is a decrease in image quality and a loss of a single stop of light.  Will this cancel out the benefit of the improved body/sensor? Would there be unacceptable increased vignettingand distortion more obviously displayed by the full frame sensor and further exacerbated by the teleconverter.  

So an interesting hypothesis to test would be:

The 5dMkII with a 1.4x teleconverter has equivalent resolution and quality to a 20/30D without teleconverter???

Off thread topic but I have the Canon 1.4 txII and there is no evidence of reduced image quality when used with a quality lens such as the 500/4. Reduced IQ with the 1.4 and 2x tc on a quality lens has everything to do with long-lens technique and very little to do with glass.
Title: Interesting Comparison
Post by: fike on December 16, 2008, 09:01:45 AM
Quote from: John Chardine
Off thread topic but I have the Canon 1.4 txII and there is no evidence of reduced image quality when used with a quality lens such as the 500/4. Reduced IQ with the 1.4 and 2x tc on a quality lens has everything to do with long-lens technique and very little to do with glass.

I have observed some decrease in sharpness when using the teleconverter.  I think my bigger problem with the 1.4 tele is that with my 100-400 f/4.5-5.6 and my 30D, it doesn't have autofocus.  That is a big issue for me, particularly when I am shooting birds or something.  With the 5DMkII body I would be able to focus with the combo though.  

Again, as someone else mentioned, I think an interesting comparison would be 5dMkII with a 400mm lens and 1.4x teleconverter compared to the 50D with the same lens and no teleconverter.
Title: Interesting Comparison
Post by: stever on December 16, 2008, 11:06:01 AM
if you have good light, the 100-400 will autofocus slowly with the last 3 pins taped on the lens and the 1.4x
Title: Interesting Comparison
Post by: fike on December 16, 2008, 03:47:30 PM
Quote from: stever
if you have good light, the 100-400 will autofocus slowly with the last 3 pins taped on the lens and the 1.4x

Yes, I have messed with that.  Unfortunately, it isn't fast enough for birds.
Title: Interesting Comparison
Post by: Ray on December 16, 2008, 04:23:17 PM
Quote from: fike
Again, as someone else mentioned, I think an interesting comparison would be 5dMkII with a 400mm lens and 1.4x teleconverter compared to the 50D with the same lens and no teleconverter.

The 5D2 with 1.4x extender, after cropping the image in post processing, would have very similar resolution to a 50D without extender. I wouldn't like to predict which image, at the extreme pixel-peeping level, would be better, but I'm fairly confident that the differences would be too small to worry about.

The problem with the Canon 100-400 IS, is that it's not at its sharpest at F5.6. (It least my copy isn't, and I believe my copy is fairly typical and also not one of the early copies). This means, when the best resolution is a priority, one feels compelled to use F8 at 400mm. With a 1.4x extender, F8 becomes F11, so right away there is a penalty...no autofous and also the need for a slower shutter speed or a higher ISO. That also means that any noise advantage the 5D2 might have (compared with the 50D without extender) is wiped out.
Title: Interesting Comparison
Post by: fike on December 16, 2008, 06:57:32 PM
Quote from: Ray
The 5D2 with 1.4x extender, after cropping the image in post processing, would have very similar resolution to a 50D without extender. I wouldn't like to predict which image, at the extreme pixel-peeping level, would be better, but I'm fairly confident that the differences would be too small to worry about.

The problem with the Canon 100-400 IS, is that it's not at its sharpest at F5.6. (It least my copy isn't, and I believe my copy is fairly typical and also not one of the early copies). This means, when the best resolution is a priority, one feels compelled to use F8 at 400mm. With a 1.4x extender, F8 becomes F11, so right away there is a penalty...no autofous and also the need for a slower shutter speed or a higher ISO. That also means that any noise advantage the 5D2 might have (compared with the 50D without extender) is wiped out.

I go a little farther. I tend to think that the 100-400 is sharpest at f/11.  Another reason why the 5dMkII quality improvements seem to be a wash when considering the use of long lenses.  But then again, f/11 is around the area where the 50D is considered diffraction limited.
Title: Interesting Comparison
Post by: Ray on December 16, 2008, 09:23:26 PM
Quote from: fike
I go a little farther. I tend to think that the 100-400 is sharpest at f/11.  Another reason why the 5dMkII quality improvements seem to be a wash when considering the use of long lenses.  But then again, f/11 is around the area where the 50D is considered diffraction limited.

I used to think my 100-400 was sharpest at F11, probably because any slight misfocussing was less apparent at F11. More careful testing revealed it was actually very marginally sharper at F8, but so close one could consider both F8 and F11 to be about equal. The real advantage of F8 is that it allows the use of a faster shutter speed which, with a telephoto lens, is often essential for a sharp result, even with IS enabled.

The fact is, the use of extenders seems to provide, at best, only marginally more detailed results. They are of dubious value in my opinion.

Here are a few 200% crops using the 100-400 with my 50D. Moving from left to right, (1) 400mm at F8 without extender, (2) F8 with extender, (3) F11 with extender, (4) F16 with extender.

The first image on the left, without extender, was cropped to the FoV of the other images and then interpolated to the same size. All images were sharpened in ACR to the same extent (amount 25, Detail 100%). As you can see, both the F11 and F16 shots with extender are about equal, but are both noticeably better than the F8 shots with and without extender.

I suppose if one were photographing an elephant at that distance, it might be worthwhile using an extender   .

[attachment=10330:Comparis...8_to_F16.jpg]
Title: Interesting Comparison
Post by: fike on December 17, 2008, 10:33:02 AM
Quote from: Ray
I used to think my 100-400 was sharpest at F11, probably because any slight misfocussing was less apparent at F11. More careful testing revealed it was actually very marginally sharper at F8, but so close one could consider both F8 and F11 to be about equal. The real advantage of F8 is that it allows the use of a faster shutter speed which, with a telephoto lens, is often essential for a sharp result, even with IS enabled.

The fact is, the use of extenders seems to provide, at best, only marginally more detailed results. They are of dubious value in my opinion.

Here are a few 200% crops using the 100-400 with my 50D. Moving from left to right, (1) 400mm at F8 without extender, (2) F8 with extender, (3) F11 with extender, (4) F16 with extender.

The first image on the left, without extender, was cropped to the FoV of the other images and then interpolated to the same size. All images were sharpened in ACR to the same extent (amount 25, Detail 100%). As you can see, both the F11 and F16 shots with extender are about equal, but are both noticeably better than the F8 shots with and without extender.

I suppose if one were photographing an elephant at that distance, it might be worthwhile using an extender   .

[attachment=10330:Comparis...8_to_F16.jpg]

This is some extreme pixel-peeping, but I will bite.  I thought the most detailed versions were the f/11 and f/16 with 1.4x teleconverter--particularly the f/11.  You are correct in saying that the difference is subtle, but if you focus on the leaves on the sapling tree, you can definitely see more information there.  When I shoot large expanses of leafy landscape, I don't like those leaves to dissolve into an amalgam of mush. I like some definition.  

Now, I still want to see the same comparison with the 5dMkII, 50D and 30D.  Does anyone have that series of cameras to work with.  Unfortunately all the online resolution samples that are done with these cameras are not applicable.  They always fill the frame with the target.  In this scenario, the 5DMkII needs to be moved farther away so that only the portion of the frame that is equivalent to the 1.6x crop factor will be visible.  This eliminates the step of resizing in photoshop or some other tool.
Title: Interesting Comparison
Post by: Ray on December 18, 2008, 08:01:57 PM
Quote from: fike
Now, I still want to see the same comparison with the 5dMkII, 50D and 30D.  Does anyone have that series of cameras to work with.  Unfortunately all the online resolution samples that are done with these cameras are not applicable.  They always fill the frame with the target.  In this scenario, the 5DMkII needs to be moved farther away so that only the portion of the frame that is equivalent to the 1.6x crop factor will be visible.  This eliminates the step of resizing in photoshop or some other tool.

It so happens that a 1.4x extender on a 5D2 produces very closely a pixel for pixel comparison with a 50D used without extender. Since the 20D pixel pitch is the same as that of the 5D2, and since the noise characteristics of the 5D2 pixel are very similar to those of the 20D, bearing in mind that at both cameras' nominated ISO values the 20D has an actual ISO which is higher, it becomes possible to simulate the performance of the 5D2 using the 20D.

Since there's a likelihood that I will eventually buy a 5D2 after the initial mad rush is over, I decided to do my own comparisons out of curiosity.

The following shots compare the 20D using a 400mm lens and 1.4x extender, with the 50D using the same 400mm lens without extender. I've used the same aperture of F11 for both shots since I didn't want to give a shutter speed advantage to either of the cameras. As it so happens, the 20D shot is at 1/800 sec as opposed to 1/640th for the 50D, despite the fact that ISO 800 was used for both shots and despite the fact that both shots are equally well exposed.

This confirms other results which indicate that the 20D is more sensitive than later Canon models, including the 5D2. However, that's all right for the purposes of these tests because a 560mm lens needs a faster shutter speed than a 400mm lens.

For those who don't want to bother looking at the following images, the deductions and conclusions are:

(1) When maximum telephoto reach is required, the 5D2 with 1.4x extender will produce detail and sharpness on a par with the 50D without extender, when images are viewd at 100%.

At an extreme pixel-peeping level, at 200% and 300% magnification, the 50D still has the edge, though, and this edge would have been slightly greater if I had used the 50D at F8.

(2) Despite the greater exposure given to the 50D shot at the same ISO, it appears to be noisier that the 20D shot, once again confirming that the smaller 50D pixel really is noisier than the 20D pixel and that total image noise is only equalled when the 50D image is downsampled to the 20D image size (or the 20D image is upsampled).

(3) At the lower edge of the 20D image, nearest the camera, it can be seen that the 20D shot is beginning to lose sharpness due to its shallower DoF. This is more apparent in the corners where the shallower DoF is combined with (I suspect) a fall-off in resolution due to the presence of the converter. In other words, the 560mm lens has poorer edge and corner resolution than the 400mm lens.

It should be noted that introducing an extender always reduces DoF, even when the main lens is at the same aperture. For example, a 400mm lens at F8 becomes a 560mm lens at F11 when a 1.4x extender is used. The DoF of a 560mm lens at F11 is still slightly shallower than a 400mm lens at F8.

[attachment=10394:Full_scene.jpg] [attachment=10395:Comparison_upper.jpg]  [attachment=10396:Comparison_middle.jpg]  [attachment=10397:Comparison_lower.jpg]  [attachment=10399:300__noi...mparison.jpg]  [attachment=10398:300__noi...midtones.jpg]
Title: Interesting Comparison
Post by: fike on December 19, 2008, 07:31:22 AM
Quote from: Ray
It so happens that a 1.4x extender on a 5D2 produces very closely a pixel for pixel comparison with a 50D used without extender. Since the 20D pixel pitch is the same as that of the 5D2, and since the noise characteristics of the 5D2 pixel are very similar to those of the 20D, bearing in mind that at both cameras' nominated ISO values the 20D has an actual ISO which is higher, it becomes possible to simulate the performance of the 5D2 using the 20D.

Since there's a likelihood that I will eventually buy a 5D2 after the initial mad rush is over, I decided to do my own comparisons out of curiosity.

The following shots compare the 20D using a 400mm lens and 1.4x extender, with the 50D using the same 400mm lens without extender. I've used the same aperture of F11 for both shots since I didn't want to give a shutter speed advantage to either of the cameras. As it so happens, the 20D shot is at 1/800 sec as opposed to 1/640th for the 50D, despite the fact that ISO 800 was used for both shots and despite the fact that both shots are equally well exposed.

This confirms other results which indicate that the 20D is more sensitive than later Canon models, including the 5D2. However, that's all right for the purposes of these tests because a 560mm lens needs a faster shutter speed than a 400mm lens.

For those who don't want to bother looking at the following images, the deductions and conclusions are:

(1) When maximum telephoto reach is required, the 5D2 with 1.4x extender will produce detail and sharpness on a par with the 50D without extender, when images are viewd at 100%.

At an extreme pixel-peeping level, at 200% and 300% magnification, the 50D still has the edge, though, and this edge would have been slightly greater if I had used the 50D at F8.

(2) Despite the greater exposure given to the 50D shot at the same ISO, it appears to be noisier that the 20D shot, once again confirming that the smaller 50D pixel really is noisier than the 20D pixel and that total image noise is only equalled when the 50D image is downsampled to the 20D image size (or the 20D image is upsampled).

(3) At the lower edge of the 20D image, nearest the camera, it can be seen that the 20D shot is beginning to lose sharpness due to its shallower DoF. This is more apparent in the corners where the shallower DoF is combined with (I suspect) a fall-off in resolution due to the presence of the converter. In other words, the 560mm lens has poorer edge and corner resolution than the 400mm lens.

It should be noted that introducing an extender always reduces DoF, even when the main lens is at the same aperture. For example, a 400mm lens at F8 becomes a 560mm lens at F11 when a 1.4x extender is used. The DoF of a 560mm lens at F11 is still slightly shallower than a 400mm lens at F8.

[attachment=10394:Full_scene.jpg] [attachment=10395:Comparison_upper.jpg]  [attachment=10396:Comparison_middle.jpg]  [attachment=10397:Comparison_lower.jpg]  [attachment=10399:300__noi...mparison.jpg]  [attachment=10398:300__noi...midtones.jpg]


That is a fantastic analysis and a very creative way to compare the cameras without having the 5DMkII to look at.  As is so often the case with pixel-peeping, there is a [email protected] of difference when the marketing would have you believe that the difference is dramatic.  

What are you waiting for on that 5dMkII?    I want to see the comparison of it to 20D.  
Title: Interesting Comparison
Post by: jani on December 19, 2008, 07:44:13 AM
Quote from: fike
What are you waiting for on that 5dMkII?    I want to see the comparison of it to 20D.  
Robert S. Blum (http://www.luiswatkins.com/5dmarkii/index.htm)
Imaging Resource (http://www.imaging-resource.com/IMCOMP/COMPS01.HTM) (select both cameras from the dropdown lists)
Title: Interesting Comparison
Post by: fike on December 19, 2008, 09:10:16 AM
Quote from: jani
Robert S. Blum (http://www.luiswatkins.com/5dmarkii/index.htm)
Imaging Resource (http://www.imaging-resource.com/IMCOMP/COMPS01.HTM) (select both cameras from the dropdown lists)


I've seen those.  They don't make the comparison we are discussing.  What we are discussing is a way to normalize the pixel pitch between the 20/30D generation and the newer 5DMkII generation.  To do this, you would need to have the resolution target on the 5dMkII samples appear smaller in the image so that it only covers the smaller area of the full frame 5dMkII sensor that is covered by an APS-C sensor.  Then, you can compare the quality of the pixels without debating the influence of pixel-pitch.  

This method of analysis is mostly important for people who are shooting telephoto.  If you care about wide angle, this analysis is less pertinent.  As you may be aware, a 400mm lens on an APS-C sensor has the same field of view as a 640mm lens would on a full frame camera.  Because we are trying to compare sensors and you can't easily compare a 400mm lens to a 640mm lens, the alternative way is to crop the 5dMkII image to have the same field of view achieved by the 20D.  This provides an elegant analysis because the pixel pitch and number of megapixels is the same at 8MP for the 20D and for the cropped 5dMkII.
Title: Interesting Comparison
Post by: Ray on December 19, 2008, 09:49:17 AM
Quote from: fike
What are you waiting for on that 5dMkII?    I want to see the comparison of it to 20D.  
 

I doubt whether I would get one before March 2009 if I put in an order today. Also, I'm not keen on being amongst the first to get a new product. There's already a 'black dots' problem which Canon are looking at, isn't there?

The fact is, I'd like a D700, an A900 and a 5D2. I can't afford all three and am totally paralysed   . (Er! with indecision).
Title: Interesting Comparison
Post by: jani on December 19, 2008, 10:42:36 AM
Quote from: fike
I've seen those.  They don't make the comparison we are discussing.  What we are discussing is a way to normalize the pixel pitch between the 20/30D generation and the newer 5DMkII generation.  To do this, you would need to have the resolution target on the 5dMkII samples appear smaller in the image so that it only covers the smaller area of the full frame 5dMkII sensor that is covered by an APS-C sensor.  Then, you can compare the quality of the pixels without debating the influence of pixel-pitch.
Essentially, what you're saying could be compressed to:

You want the tester to use the same distance to the chart with both cameras.

So what you need to do, is to get someone with a 5D MkII to print the test chart used for the EOS 20D at imaging-resource.com and photograph it at the same distance as they did for the 20D, under similar lighting conditions.

That ought to be possible.

I only have the 20D and don't plan on buying the 5D MkII, so I can't really help.
Title: Interesting Comparison
Post by: fike on December 19, 2008, 11:00:17 AM
Quote from: jani
Essentially, what you're saying could be compressed to:

You want the tester to use the same distance to the chart with both cameras.

So what you need to do, is to get someone with a 5D MkII to print the test chart used for the EOS 20D at imaging-resource.com and photograph it at the same distance as they did for the 20D, under similar lighting conditions.

That ought to be possible.

I only have the 20D and don't plan on buying the 5D MkII, so I can't really help.

Yes, same lens, same distance.  

Determine the distance using the 20/30D and then use the same distance and lens with the 5dMkII.  I am kind of surprised none of the reviewers have thought to do this analysis.  If you really appreciate the added reach that a cropped sensor provides, this really gives you an apples to apples comparison that includes lens range as a factor.
Title: Interesting Comparison
Post by: Ray on December 19, 2008, 11:08:12 AM
Quote from: jani
Essentially, what you're saying could be compressed to:

You want the tester to use the same distance to the chart with both cameras.

So what you need to do, is to get someone with a 5D MkII to print the test chart used for the EOS 20D at imaging-resource.com and photograph it at the same distance as they did for the 20D, under similar lighting conditions.

That ought to be possible.

I only have the 20D and don't plan on buying the 5D MkII, so I can't really help.

If you use the same lens on both cameras, shoot from the same distance and compare equal size crops, the crops will fairly exactly be comprised of the same number of pixels and you will, in fact, be able to compare noise and resolution at the pixel level without introducing interpolation effects.

However, the issue for me, and I think for fike also, is the possible advantage of carrying a 50D as a second camera for long telephoto work. Carrying a 1.4x extender with the 5D2 could be just as effective, except perhaps for the loss of autofocussing that use of an extender entails with a lens like the Canon 100-400 IS.

On the other hand, if you were to use the 1.4x extender with the 50D instead of the 5D2, you could expect to get a marginally sharper and more detailed result. But just how significant that extra detail would be in practice, or on print, is another matter. I wish I owned a top notch telephoto prime to carry out such comparisons, but unfortunately I don't.