Question on TC effectivness

[tiziano]

Question on TC effectivness.

I was wondering… on the sensor of my 20D, I have something like 155 pixels for linear mm, which will be translated in about 77 lines/mm. Now, given the performances of most lenses, this implies that, if the man behind the camera doesn’t mess up, the bottleneck for image definition is the lens, and not the sensor.
Now, if this above is correct, assuming that my 1.4X or 2X TC is attached at the rear of my lens, I image that, provided that the additional lenses of the TC don’t degrade the IQ (optimistic assumption) the TC is enlarging the imaging coming out of the lens by 1.4 or 2 times, therefore diminishing the number of L/mm by the same factor.
My conclusion would then be that the TC is not bringing any benefit, compared to just shooting without TC and then enlarging your final image in Photoshop. This would not be the true, if the resolving capabilities of the lens were higher then these of the sensor.
I am sure I am making a mistake somewhere; otherwise TC would not be so extensively used. But, can someone tell were I am wrong?
Thanks
Tiziano

[CyberDyneSystems]

What lens?

That's where you may be going wrong.. the lenses that TCs work best with not surprisingly have resolution to spare.

Put a 1.4x behind some tele zooms and indeed the results are no good (for many reasons possibly but perhaps your theory is part of it?)

Put a 1.4x behind a lens that has much more resolution due to it's dinner plate sized elements,. and it's a totally different story.

[tiziano]

Well, I would say all...
Here below the MTF of the famous Canon 200 f/1.8, one of the sharpest tele of all times, and the famous 70-200 f2.8, to which everybody attaches a TC.
Please notice that the 77 L/mm I gave above, are actually lines-pairs /mm.
The two lenses above perform, at their best at around 30~35 lines-pairs /mm.
http://www.wlcastleman.com/equip/reviews/200mm/index.htm

[pieq314]

Well, I would say all...
Here below the MTF of the famous Canon 200 f/1.8, one of the sharpest tele of all times, and the famous 70-200 f2.8, to which everybody attaches a TC.
Please notice that the 77 L/mm I gave above, are actually lines-pairs /mm.
The two lenses above perform, at their best at aroun 30~35 lines-pairs /mm.
http://www.wlcastleman.com/equip/reviews/200mm/index.htm
My memory+understanding is that a telephoto converter (especially 1.4x) does not degrade the lens significantly. That is, if a lens has 35 line pairs per mm, after attaching the 1.4x converter, it will have a resolution of close to the original 35 line paris/mm, not 35/1.4=25 line pairs/mm. So your camera will see more details of the subject.

[tiziano]

My memory+understanding is that a telephoto converter (especially 1.4x) does not degrade the lens significantly. That is, if a lens has 35 line pairs per mm, after attaching the 1.4x converter, it will have a resolution of close to the original 35 line paris/mm, not 35/1.4=25 line pairs/mm. So your camera will see more details of the subject.

Sorry, I believe this is physically impossible. keeping the same amount of details per square mm would mean, on an enlarged image, to increment the number of details. The TC would be providing information it didn't receive...
The TC increases the focal lenght, and therefore projects a smaller angle of view on the same area, the sensor. You will get an enlarged central area, and the external area cropped.

[pieq314]

Not really. The TC becomes part of the lens. It is not re-imaging the image by the original lens.

Here is what would happen to the way you predicted: place a ground glass on the image plane (where the sensor would be), place the TC after that and look at the image formed by the TC. Yes, the final image cannot be better than the image at the ground glass (by the original lens), that is, you just get a bigger image without additional details.

But here is what actually happens. The TC is part of the lens now. It is not magnifying an image, but functions just as additional elements of a "new" lens, so the image could be theoretically better in two ways:

1. Resolve more details in the subject (by a factor of a little less than 1.4x for 1.4x converter). That is, if measured at the sensor, the lines/mm does not change much.

2. It could theoretically increase the lines/mm at the sensor. (Practically, I know it is impossible, but I just want to bring up the point). Think it this way, the original lens has probably over 10 elements. If you take off the last 5 elements, do you get a better resolution at the sensor(ignore the fact that the focus length is different)? The answer is NO. What this means is that the last 5 elements are improving the lens. This brings us back the the teleconverter: it does not violate any physics for not downgrading the lines/mm at the sensor, at least not by a factor of 1.4.

[Wilt]

Sorry, I believe this is physically impossible. keeping the same amount of details per square mm would mean, on an enlarged image, to increment the number of details. The TC would be providing information it didn't receive...
The TC increases the focal lenght, and therefore projects a smaller angle of view on the same area, the sensor. You will get an enlarged central area, and the external area cropped.

tizano, if you look at the 70-200 on photozone.de, you see the following at 200mm vs. 280mm (with 1.4x)...

MTF with lens wide open:
1820 center at 200, 1805 center at 280.
1747 edge at 200, 1617 edge at 280.

Center drops less than 1%...kinda blows a hole in the belief that resolution performance of the lens has to suffer when you mount an extender!

[pieq314]

tizano, if you look at the 70-200 on photozone.de, you see the following at 200mm vs. 280mm (with 1.4x)...

MTF with lens wide open:
1820 center at 200, 1805 center at 280.
1747 edge at 200, 1617 edge at 280.
Now I remember where I saw the 1.4x test. I tried but failed to recall where I saw it.

So I just took another look. At 200mm f/2.8, center resolution is 1749. After attaching 1.4x, it becomes 280mm f/4, with a center resolution of 1728. So the resolution barely changes with the 1.4x on.

[Wilt]

So we don't confuse tiziano by two different results on the 70-200...my message was about the f/4L lens, pieq314 was about the f/2.8L

[Woolburr]

Looking at charts is seldom the most effective method of testing the effectiveness of using or not using a TC....The laws of physics dictate what is going to happen when you utilize a TC....what it does not dictate is how well the final image translates to viewing. You are always going to have a loss of image quality regardless if you add a TC or if you add a filter...to some degree, anything you place in front of or behind the original lens is going to degrade the image. The question becomes at what point does it become visually unacceptable? The only way of knowing is not by looking at charts or calculating lines per mm, but by taking real world images and asking people if they still find the image to be appealing to the eye.

[tiziano]

tizano, if you look at the 70-200 on photozone.de, you see the following at 200mm vs. 280mm (with 1.4x)...

MTF with lens wide open:
1820 center at 200, 1805 center at 280.
1747 edge at 200, 1617 edge at 280.

Center drops less than 1%...kinda blows a hole in the belief that resolution performance of the lens has to suffer when you mount an extender!

Now I remember where I saw the 1.4x test. I tried but failed to recall where I saw it.

So I just took another look. At 200mm f/2.8, center resolution is 1749. After attaching 1.4x, it becomes 280mm f/4, with a center resolution of 1728. So the resolution barely changes with the 1.4x on.


Thanks Wilt and Pieq for the explanations. I have given a look at the photozone.de site, and at the Imatest site that explains what are these numbers:
"1820 center at 200, 1805 center at 280. "
My understanding is that 1820 and 1805 LW/PH are the overall number of: “line widths per picture height”. This is actually the number of horizontal lines that will come out of the lens, once you have totally filled the sensor with the image with the test chart.

From photozone.de FAQ:
These test charts have a size of 120x80cm. The chart should cover the complete image field of a lens. This is regardless of the lens type (ultra wide, tele, macro ...) so the object magnification (of the test chart) remains always the same. For an APS-C DSLR this translates roughly to a camera-to-chart distance of 60x focal length (for full frame it would be 40x focal length) - e.g. 18m for a 300mm lens.

Therefore, after this test is executed, the Imatest algorithm calculates the maximum number or lines that the lens can resolve, which in this case 1820 at fl 200mm without TC, 1805 at fl 280 with TC.
Ok, so far we know that the TC has not degraded the image in a meaningful way.
But, my question is: has the TC increased the amount of details we can capture?
We know now that, with or without TC, the lens can resolve about 1820 horizontal lines on a 35 mm film, or on a full frame sensor, both 36x24 mm.
1820/24 is the maximum numbers of lines per mm that will be resolved, and 1820/24=75 (about). Same number provided by the wlcastleman web site. Please notice that our 20D sensor still resolves 155 lines per mm.
Conclusion: so far we have proven that wlcastleman and photozone.de provide reliable testing, and that the TC does not affect IQ.
But still don’t see the TC to have increased the number of captured details, when used on a 20D.

Sorry, maybe it is just me being dummy…

EDIT: I'm sure I'm wrong somewhere. I guess I'll better go to sleep. Tomorrow I'll probably understand. Thanks for your time. :)

Edit 2: I got it. If the lens can resolve the same details at a higher distance, I have achieved my original purpose of extending properly the focal lenght.
I think I'll sleep better now... :)
Thanks again!!!

[Wilt]

Conclusion: so far we have proven that wlcastleman and photozone.de provide reliable testing, and that the TC does not affect IQ.
But still don’t see the TC to have increased the number of captured details, when used on a 20D.

What you are missing is that a given amount of subject is magnified 1.4x, so it occupies 1.4x more millimeters. so instead of filling 100 mm it fills 140 mm and you get about 1800 ll/mm for both, so 180000 lines vs. 252000 lines to represent the same length of subject. More detail.

[tiziano]

Thanks Wilt!
I was just figuring that out! About time... :o