RF 100-500 worse with RF 1.4x?

[Frodo]

Hello friends (especially AlanF and usern4cr)

I've been following threads comparing the RF 100-500 with and without the RF 1.4x, with the RF 800/11, as well as threads elsewhere. Shooting from the same distance with an R5 I expected that the resolution of the RF 100-500 would be better with the RF 1.4x. I just received a new RF 1.4x and after shooting a few tests, I was surprised to find the RF 100-500 performed better when cropped than with the RF 1.4x. I wonder if there is a problem with my testing or if my RF 1.4x is somehow substandard. Your thoughts would be appreciated.
I shot a test chart and a toy penguin at about 10m in bright sunshine, the test chart on a sturdy carbon fibre tripod with 2 second delay with manual focusing; the penguin by hand kneeling ont the ground using EFCS and eye tracking (a real world situation) . My first trial using electronic first curtain shutter suggested possible vibration, so I repeated the test chart photos with electronic shutter. The photos were both processed in DXO Pure Raw2.
Comparing the test chart in Lightroom with the 500mm photo at 150% and the 700mm photo at 107% showed the 700mm photo to be softer. This improved when I added quite a bit of clarity.
The penguin photos when viewed at 196% and 143% (for the 500mm and 700mm) showed similar levels of detail.
The first photo shows the test target and penguin at 500mm uncropped.
The second shows the test target viewed at 150 and 107%.
The third photo as with #2 but added clarity in the 700mm photo.
The final photo shows the penguins viewed at 196% and 143% with no additional editing.
Your thoughts please!

 

[Frodo]

I ran some more tests and concluded that vibration is a major factor to be considered when running tests of long focal lenses. I tried to delete this thread, but couldn't - perhaps my learnings will be of use. This analysis is not academic as I regularly print my landscape images 30 inches wide and would like the option to also print large wildlife photos.
Firstly as AlanF mentioned in another thread, play close attention to the fine concentric circles in a test target. These circles make vibration quite evident, where there is blurring on opposite sides of the circle, but not perpendicular. I ran a series of photos at 500mm and 700mm, with my camera on my tripod (a sturdy Manfrotto carbon tripod). I concluded that at shutter speeds twice the focal length (i.e. 1/1000 and 1/1500) I could count on only 1 or 2 photos out of 5 being sufficiently sharp to make a valid reference. Secondly, I didn't see a lot of difference between ECFS and electronic shutter. Thirdly, IS definitely makes things worse when on a tripod at shutter speeds twice the focal length.
I shot photos at about 30m away, and bearing the previous comments in mind, I conclude:
- the naked 100-500 is crisper and sharper than the lens with the 1.4x, provided that viewing is less than 100%
- once the photo starts to get a little pixelated, the 1.4x provides a higher resolution image - see the first photo.
What if the 500mm photo is upscaled 2x with Gigapixel AI? The second photo shows the difference ('edit" means upscaled). I don't think there is much in it, noting that there is still some vibration apparent.
I might repeat this at some point using higher shutter speeds and multiple photos.
So to answer my question, my technique is/was weaker than the RF 1.4x. This also shows how important technique is in the real world when seeking sharp images with long lenses,

 

[Frodo]

What a difference proper technique makes! I set up the target at 10m in the evening and used flash (which completely overpowered the ambient light).
The first photo shows a dramatic improvement in the sharpness of both images, especially the 700mm (100% for the 700mm image and 136% for the 500mm to make it the same size on screen), Line pairs are visible in the 8.0 group, whereas the 500mm resolves to the 7.1 group.
The second photo sets the 500mm photo at 100% and the 700mm at 143%. It shows that Gigapixel adds contrast and a few artifacts, but no fine details. The contrast of the 700mm images could be increased without artifacts.
I conclude that the RF 1.4x does indeed add some resolution.
I wish I had done this comparison first!

 

[koenkooi]

This reminds me a bit of the R5 400MP multi shot thing, it showed how wobbly regular camera use really is.

 

[neuroanatomist]

vibration is a major factor to be considered when running tests of long focal lenses

Indeed. And when using them. It’s why I have an RRS TVC-33 tripod for my 600/4. Even then, for example if testing the lens at home I need to set up in the basement where there’s a concrete slab under the flooring, because the wooden subfloors at higher levels of the house transmit too much vibration.

Another factor is image stabilization – it takes ~0.5 s for lens IS to fully activate, and while its ’spinning up’ it makes IQ worse. So if you just smoothly press the shutter in one motion, IS will hurt not help.

 

[AlanF]

What a difference proper technique makes! I set up the target at 10m in the evening and used flash (which completely overpowered the ambient light).
The first photo shows a dramatic improvement in the sharpness of both images, especially the 700mm (100% for the 700mm image and 136% for the 500mm to make it the same size on screen), Line pairs are visible in the 8.0 group, whereas the 500mm resolves to the 7.1 group.
The second photo sets the 500mm photo at 100% and the 700mm at 143%. It shows that Gigapixel adds contrast and a few artifacts, but no fine details. The contrast of the 700mm images could be increased without artifacts.
I conclude that the RF 1.4x does indeed add some resolution.
I wish I had done this comparison first!

You are a man after my own heart using Bob Atkins charts - they are for me the best. I thoroughly agree with your analysis all the way through. Gigapixel never increases resolution of parallel lines or concentric circles on chart - in fact often the opposite. Where it does appear to is in real life situations where say the feathers of birds seem to be brought out. Staring at your good chart under flash, it seems to me that the 9 resolution with the 1.4 at 700mm is matched by the bare 500 between the 7.1 and 8, which gives the 1.4x about an effective 1.2xTC. In general, sharpening increases contrast but not real resolution.

 

[Frodo]

Looking more closely at my last post, I wondered if the actual resolution of the test chart was insufficient to distinguish between the RF 100-500 without and with the RF 1.4x. So I reshot the test chart at 20m, illuminated with a remote flash at ISO 400 - ambient illumination is negligible - none of the images indicated movement blur.. The files were imported into Lightroom, the white balance normalised at a consistent point and the exposure increased so that the whites were reading 90%. Screenshots were taken of paired images, with the higher res image at 100% and the lower res image at 143% (LR wouldn't allow 140%).
#1: The 700mm image at 5.0 is similar to the 500mm image at 4.0. The 700mm image has less local contrast.
#2: Adding 100% texture to the 700m produces similar contrast to the 500mm, but no increase in resolution
#3: I also moved the camera closer so that the chart at 500mm is similar to that of the 700mm. The 500mm resolves to 5.6 on the chart compared to 5.0 for the 700mm. The 700mm image has lower contrast.
#4: Converting the RAW file with DXO Pure Raw 2 significantly improves contrast with the 700mm
What does this mean in practice? I photographed some terns this morning, at 500mm and 700mm at the same distance.
#5: In good light, 700mm does resolve more detail than 500mm.
#6: The 500mm image is now upscaled with Gigapixel AI and sized on the screen to the 700mm at 100%. GP AI adds contrast, but not much in the way of fine feather detail.
I conclude that, in good light and good technique, the RF 1.4x adds noticeable resolution of subjects photographed at the same distance with the bare lens.