Forgot Password
Pentax Camera Forums Home
 

Reply
Show Printable Version Search this Thread
01-12-2009, 11:18 AM   #46
Pentaxian
Marc Sabatella's Avatar

Join Date: Sep 2006
Location: Denver, CO
Photos: Gallery
Posts: 10,686
QuoteOriginally posted by Class A Quote
Do you find my tests as inconclusive as Dan perceives them?
Maybe not *as* inconclusive, but I do admit he has raised some legitimate doubt (about the test, not about the conclusion). When you first posted the test, I looked at the text within the pictures and thought you had clearly shown the 50 was sharper (as well all "know" it is) - end of story. But then when Dan pointed out the texture in the area around the sail, I looked harder, and of course he has a point, so I'm confused.

So I'd ask you: *is* that moire or is it real detail? That is, do you see that text on the actual subject? If it is there, it seems odd to me the 50 would show more detail in one area but less in another, which would lead me to suspect either a lens defect or else a DOF issue (since the "plane" of focus is not actually flat). Also - how perpendicular to the subject were you? If that wasn't precise, then of course different parts of the subejct would be at different distances.

As things stand, I have to agree with one of Dan's points: your tests are not as conclusive as they first appear. However, I do *not* agree with his basic premise that 6MP isn't enough to show the difference a prime can make. I know from my own experience that 6MP is *more* than enough to show the difference in the lenses *I* compared, and the sharpness in *other* areas of your picture *appear* consistent with that. But given that that your results do not appear to hold across the board, I don't think we can rule out some sort of testing error.

QuoteQuote:
Note that the 18-250 is not the 18-55. It is quite possible that the former is a bit sharper around the centre.
Possible, indeed - especially at that focal length and aperture. The 18-55 is wide open at f/5.6 at 50-55mm, so it suffers accordingly. The 18-250 is, I assume, at least half a stop from wide open at those focal lengths?

01-12-2009, 02:02 PM   #47
Pentaxian
Class A's Avatar

Join Date: Aug 2008
Location: Wellington, New Zealand
Posts: 9,063
QuoteOriginally posted by audiobomber Quote
- You cannot compare resolution in your photos with mine, for all the reasons I stated previously.
I can clearly observe though that the sharpness of your K100DS images is nowhere near it should be. You have used the "bright" setting which does more sharpening than my "natural" setting. Still both your K100DS images look fuzzy.

I have downloaded both of your K100DS images and flicked back and forth between them in Picasa. Guess what:
  1. They are taken from a different position and show a different field of view. Difficult to make a resolution comparison under such circumstances.
  2. The kit lens image looks slightly better than the prime image (do the flick back and forth comparison and watch the structure of the second cereal ring under the beginning of the big "B"). This suggests that the prime hasn't reached its potential (probably due to AF problems). The prime should at least be as good as the kit lens, even if your "6MP are outresolved" theory were true.
  3. In neither of the images I see the contrast gradients I would expect to see from a K100DS and that would begin to allow a comparison of resolution.
Summarising, your K100DS images are just not fit for making any conclusions about resolution. Independently of the quality of my tests, I think that's a fair statement to make and one way of avoiding running in circles, would have been to acknowledge that.

QuoteOriginally posted by audiobomber Quote
- I have checked the focus on both my cameras using a focus chart with three primes and a zoom, under tungsten light and daylight. Both cameras performed equally well. AF and MF were spot on at alll apertures in daylight with each lens.
Wow, that's rare, I'd say. But assuming it is the case then obviously
  1. The AF result for your K100DS prime shot seems to suffer from slight misfocus. It is normal for the AF system to work within tolerances and not get it a 100% right sometimes. That's why shooting flat targets without shooting a series of images with varying focus, and then picking the best one, is very dangerous.
  2. If your AF is spot on for your camera<->focus chart distance, it doesn't mean it will be spot on for other distances. The recommendation is to calibrate/check with your "expected/normal" shooting distance. I'd be surprised if your focus chart and Cheerios box distances were the same.
  3. something else is working as a low pass filter on your images. Perhaps its the K100DS JPEG output, or your rescaling, most likely misfocus, though. See my list above again.

QuoteOriginally posted by audiobomber Quote
- You are struggling to show the differences in resolution between a superzoom and a prime
The quality difference in the shots is blindingly obvious. Again, you could say it is because the superzoom frame degrades towards the corners whereas the prime doesn't and I'm not showing a resolution difference in the centre. That I could accept to some extent. And you are right: I should not have adopted your methodology (flat subject without guaranteeing optimal focus, rescaling from 1300 to 1200). I thought I'd give it a quick spin and then had the impression the results speak for themselves. My apologies to everyone for not providing even more convincing results.

QuoteOriginally posted by audiobomber Quote
I have tested these very lenses on a K20D, and I know there are major resolution differences.
We have two issues here:
  1. Your claim that the K20D clearly shows resolution differences between lenses.
  2. Your claim that the K100DS doesn't show resolution differences between the kit lens and a prime
I'm not disputing your first claim. Having said that, I don't think that your K20D images are proof of it. But that's not why I have contributed to this thread.

I'm not even opposing to the idea that the K20D, in principle, can show lens differences more easily then the K100D.

I take issue with your second claim. Your kit lens shot differs from your prime shot in perspective but if anything is slightly better. That should not be case independently of whether or not your theory is true.

QuoteOriginally posted by audiobomber Quote
There's no detailed analysis needed or excuses about maybe the prime was misfocussed.
Why do you write "excuses"?
Misfocus is an explanation as to why the prime lens didn't reach its potential in my first sequence (and probably hasn't reached it in the second either).

QuoteOriginally posted by audiobomber Quote
- Mounted on a K20D, the 18-55 has more resolution at 50mm f5.6 than the FA 50mm on a K100DS.
We can make all sorts of claims we then don't support in a valid manner.

QuoteOriginally posted by audiobomber Quote
It is still worthwhile for an owner of a 6mp camera to buy a good prime for speed, improved bokeh and colour rendition.
Agreed, but I don't understand why you don't add "image quality". Although I have failed to unequivocally demonstrate a centre sharpness superiority of the prime, I clearly demonstrated how much better the prime image is across the frame (and my crops don't even show the full image). For that reason alone it is worth using, even exotic primes on a 6MP sensor camera.

The problem with our disucssion is not that the horse is dead but that we are not beating the right way. I pointed out problems with your beating and you are not convinced by my beating. That doesn't mean the horse is dead, in other words, your theory has not been proven.

I don't have time to press this on right now, but eventually I will attempt to find optimal AF adjustments for all my lenses. This will take a lot of time since it is very easy to get wrong results. I'm preparing my own focus chart with a non-tilted target, for instance.

When I'm done with this, I'll probably try once more to investigate resolution differences between lenses using a 6MP sensor camera. I may another methodology because I believe yours is prone to deliver unreliable results, or at least, I would do series of shots with differing focus (best implemented by slightly varying camera<->subject distance between shots).

It is always a good idea to agree to disagree when discussion no longer yields progress. I just don't understand how you can go away remaining convinced that anything has been proven. What I showed with my series that a 6MP camera can show that a prime delivers better IQ across the frame then a superzoom (surprise, surprise). I'm not saying anything about centre resolution since figuring that one out requires more effort than either of us has put in so far.

Thanks for reading.

Last edited by Class A; 01-12-2009 at 03:45 PM. Reason: typos
01-12-2009, 03:20 PM   #48
Pentaxian
Class A's Avatar

Join Date: Aug 2008
Location: Wellington, New Zealand
Posts: 9,063
QuoteOriginally posted by kenyee Quote
FWIW, I think audiobomber is right. [...] After some testing this weekend, the K20D seems to show which lenses are sharper and which aren't more easily...
I'm happy with this part of Dan's statements. That doesn't make him right about resolving power of 6MP sensors, though.

QuoteOriginally posted by Marc Sabatella Quote
But then when Dan pointed out the texture in the area around the sail, I looked harder, and of course he has a point, so I'm confused.
He had a point about my first prime image. It was too much out of focus and therefore only showed better performance across the image but not in terms of centre resolution.

This is why I followed up with a second comparison in which the prime was in better focus.


QuoteOriginally posted by Marc Sabatella Quote
So I'd ask you: *is* that moire or is it real detail?
What the superzoom shows is a kind of odd moire. The package print is made of the standards circle and dots print pattern. The superzoom doesn't show this pattern just some interaction with it and the sensor / demosaicing algorithm, I think.

I have attached an image that perhaps I should not have attached. These are crops of crops, the latter of which have been downscaled. Also, what you see is at least the 3rd generation of saving with JPEG, which doesn't help to retain detail. A better comparison would have been possible if I had taken the crops out of the original images. But I want to avoid arguments about pulling a third rabbit out of a hat so I just did what everyone can do as well:
  1. get a crop of the superzoom image that Dan said to be sharper ("Sorry man, in the first set the 18-250 was sharper.").
  2. get a crop of the second prime image I provided. This one had better focus but probably still not optimal focus.
  3. put the crops next to each other.
As you can see the large pattern-like structure in the superzoom is not detail.

BTW, if you do the same with the second superzoom image (which Dan thought was less sharp) then the difference isn't as big, since the superzoom also shows more of the real detail and less of the artificial pattern. However, the difference is still there and the prime is clearly sharper. I picked the first one in order to comply with Dan's assessment. Please, anyone feel free and do the comparison with the same crop of the other superzoom picture.


QuoteOriginally posted by Marc Sabatella Quote
Also - how perpendicular to the subject were you?
I didn't put much effort in the alignment in the first series. In the second, I used a mirror to align the platform and then aligned the box with the platform.

QuoteOriginally posted by Marc Sabatella Quote
As things stand, I have to agree with one of Dan's points: your tests are not as conclusive as they first appear.
Please let me know if you still think that after having looked at the crop comparison.

EDIT: Please note that the images I attached in the thread were ~2700 pixel wide crops downscaled to 800 pixels width for the forum. That's almost a reduction of 3.4 in just one dimension. This has to be considered when judging the differences. Dan used 1300 crops and displayed them at 1200 pixel width. Any differences would be more pronounced (and still they don't show in his K100DS examples). The small crops below are not new 1:1 crops. They are just taken from my originally posted images where the differences have already been levelled somewhat by the reduction in size.

QuoteOriginally posted by Marc Sabatella Quote
The 18-250 is, I assume, at least half a stop from wide open at those focal lengths?
At 50mm (or 55mm where I used it to get near effective 50mm) it nominally shows f/4.5, so yes, f/5.6 means I stopped it down half a stop.

Warning: The following crops are crops of rescaled crops and at least the 3rd generation of JPEG. Their quality isn't as it should be. I think they still speak for themselves. Or maybe not...
First the superzoom, then the prime.

Last edited by Class A; 08-07-2010 at 09:21 AM.
01-12-2009, 07:09 PM   #49
Pentaxian
Marc Sabatella's Avatar

Join Date: Sep 2006
Location: Denver, CO
Photos: Gallery
Posts: 10,686
QuoteOriginally posted by Class A Quote
Warning: The following crops are crops of rescaled crops and at least the 3rd generation of JPEG. Their quality isn't as it should be. I think they still speak for themselves. Or maybe not...
They do for me. They clearly show that the "detail" that appeared more prominent in the superzoom shot was not detail at all. The text in the original prime shot was clearly sharper than that in the original superzoom shot. The shots below demonstrate that the prime really is reproducing the sky area more accurately, too. All in all, there is absolutely no question whatsoever left in my mind that your shots do successfully show that the 50 is sharper than the 18-250 at f/5.6 (which of course we all knew), and this is visible even on a 6MP camera (which was the only point of contention).

01-13-2009, 06:05 AM   #50
Pentaxian
audiobomber's Avatar

Join Date: May 2008
Location: Sudbury, Ontario
Photos: Albums
Posts: 6,644
Original Poster
Alright Class A, I'll concede partly.

Your photos do show a difference between the superzoom and 50mm. My 50mm photo taken on the 6mp camera is a bit off and does not properly represent the 50mm's resolution.

But I still say the 6mp camera does not show the magnitude of difference between these lenses. The thing to to do now is to repeat the test with both cameras, but I don't have time. I'm in a crunch period at work until the end of February and we're preparing our house for sale right now ("staging", as my wife calls it, which involves a lot of painting, decorating, repairing, changing light fixtures and shuffling furniture).
01-13-2009, 09:32 AM   #51
Pentaxian
Marc Sabatella's Avatar

Join Date: Sep 2006
Location: Denver, CO
Photos: Gallery
Posts: 10,686
QuoteOriginally posted by audiobomber Quote
Alright Class A, I'll concede partly.

Your photos do show a difference between the superzoom and 50mm. My 50mm photo taken on the 6mp camera is a bit off and does not properly represent the 50mm's resolution.

But I still say the 6mp camera does not show the magnitude of difference between these lenses.
I have no problem agreeing with that. The 50 is outresolving the 6MP sensor and superzoom is not. So indeed, the advantage of the 50 *is* being limited by the sensor. Not eliminated, but limited.
01-13-2009, 02:54 PM   #52
Pentaxian
Class A's Avatar

Join Date: Aug 2008
Location: Wellington, New Zealand
Posts: 9,063
Dan,

thanks for chewing through my lengthy postings.

QuoteOriginally posted by audiobomber Quote
But I still say the 6mp camera does not show the magnitude of difference between these lenses.
Possibly, probably even likely. You could be right there but certainly my tests weren't good enough to allow such a conclusion (because I think the difference is bigger than I managed to demonstrate).

QuoteOriginally posted by audiobomber Quote
we're preparing our house for sale right now
Good luck with that! I assume you will be moving and hope you'll like your new place. It'll be fun to meet here again, once we both had a chance to go back and do better tests. In my case, the objective will be mainly to get the optimal adjustment for my FA 50/1.4 and other lenses, but I will post another comparison if the results seem worth it.
01-14-2009, 06:57 AM   #53
Veteran Member




Join Date: Nov 2006
Location: Hong Kong
Posts: 1,934
Your test just verifies the trivial.

Sorry, but that's just a too trivial thing.

When either the lens or the sensor has far less resolution which is the limiting factor, the bottle neck will be shown up.

To get both highest resolution with lowest noise and most resolution out of a lens. A full frame lens on a full frame sensor will match and is the only way to go.

01-14-2009, 12:48 PM   #54
Pentaxian
Class A's Avatar

Join Date: Aug 2008
Location: Wellington, New Zealand
Posts: 9,063
QuoteOriginally posted by RiceHigh Quote
Sorry, but that's just a too trivial thing.
Not sure which test you are referring to but assuming it is mine then I'd say "yes and no".

Yes, I didn't intend my test to show something surprising.

QuoteOriginally posted by RiceHigh Quote
When either the lens or the sensor has far less resolution which is the limiting factor, the bottle neck will be shown up.
This is where the "No" comes in. From Dan's original experiment it wasn't clear whether the sensor was a bottleneck or not. So while your statement is trivially true, it wasn't clear as to whether a 6MP sensor is a bottleneck for a kit lens / superzoom.

BTW, I tried to look at this problem mathematically taking into account the sensor resolution (2008 vertical pixels) and lens MTF-50 figures of lenses (which refer to line-widths). I don't find it easy though to compare these two resolution concepts. One would have to know how many lines per millimetre (L/mm = 2 LW/PH * Height of Picture in mm) lead to a 50% relative contrast loss between the lines when viewed through 2008 LW/PH. Clearly it would have to be more than 2008 LW/PH because these could be shown with 100% contrast. I'm assuming 1.3 x as many LW will cause a 50% contrast loss. That would mean a 2008 pixel sensor has a corresponding MTF-50 figure of 2677 LW/PH. This assumes a monochrome image. Coloured images are more difficult to argue about because of the Bayer array arrangement of pixels. Just halving the sensors resolution because of the bayer array would be too crude since for instance interpolation of green pixels can help to restore detail for the image positions covered by red pixels as long as was some green component to the light hitting the red pixel and vice versa. I'll try to find about more regarding this way of doing the comparison.


QuoteOriginally posted by RiceHigh Quote
To get both highest resolution with lowest noise and most resolution out of a lens. A full frame lens on a full frame sensor will match and is the only way to go.
You cannot make the statement like that. Resolution depends on pixel density not on sensor size. So a high pixel density APS-C sensor can outresolve a low pixel density FF sensor. Furthermore full frame lenses are more difficult to get right in terms of corner performance. Someone's signature says something like "FF, yes at last we get blurred corners again, but luckily we cannot see them well because of the vignetting". While a FF system, in principle, can do everything an APS-C system can plus more, it does not automatically need to be better.

Last edited by Class A; 01-14-2009 at 11:24 PM. Reason: Changed x 3 to x 1.3
Reply

Bookmarks
  • Submit Thread to Facebook Facebook
  • Submit Thread to Twitter Twitter
  • Submit Thread to Digg Digg
Tags - Make this thread easier to find by adding keywords to it!
18-55mm, 50mm, camera, dslr, k100ds, k20d, kit, lens, photography, resolution, shot, test, time
Thread Tools Search this Thread
Search this Thread:

Advanced Search


Similar Threads
Thread Thread Starter Forum Replies Last Post
For Sale - Sold: K100DS Like New waterhouse8800 Sold Items 2 07-28-2009 08:09 PM
K100DS and Manual WB res3567 Pentax DSLR Discussion 2 05-24-2009 09:14 PM
My K100ds is broken pasipasi Pentax DSLR Discussion 5 10-06-2008 12:28 PM
K100DS Firmware aamir515 Photographic Technique 4 10-04-2008 08:00 PM
K100DS issue DanLoc78 Photographic Technique 6 05-30-2008 08:18 AM



All times are GMT -7. The time now is 05:28 PM. | See also: NikonForums.com, CanonForums.com part of our network of photo forums!
  • Red (Default)
  • Green
  • Gray
  • Dark
  • Dark Yellow
  • Dark Blue
  • Old Red
  • Old Green
  • Old Gray
  • Dial-Up Style
Hello! It's great to see you back on the forum! Have you considered joining the community?
register
Creating a FREE ACCOUNT takes under a minute, removes ads, and lets you post! [Dismiss]
Top