RF 100-500mm with RF Extender 2x

[colinrb]

I have recently used the RF 100-500mm and RF Extender 2x combination on an R5 Body and find it is pointless adding the 2x Extender.
I can get better quality by increasing the focal length in Photoshop with the lens alone at 500mm f7.1 or 500mm f8.
I am guessing this is due to diffraction as I have tested this with several lenses and diffraction starts to occur at f11 with the R5 when viewing at 100%, but is not visible on a 30" monitor with the full frame.
It becomes obvious at f16 on the full image on this monitor.
This extender is designed to work with the 100-500mm, 600 f11 and 800 f11, and as it would make these last two lenses f22, severe softening would occur, again making the extender pointless.
Has anybody else come across these issues with the 2x extender.

 

[kaihp]

@AlanF says he's blowing hot & cold on using the extenders. Not sure if it's related to the amount of light at the scene

 

[danfaz]

Has anybody else come across these issues with the 2x extender.

I have not, mainly from what I've seen and heard from other members.
The 1.4X, however, does pretty good. There is only a little reduction in quality, using it on R5 and R7 with the RF 100-500.

 

[neuroanatomist]

There’s also a sensor resolution component. The RF 2x does well on my R3, especially with the 100-300/2.8. With my EF 600/4 II, I found the EF 2xIII to be better than the RF 2x, while the respective 1.4x TCs were similar.

 

[AlanF]

In theory and pretty well in practice; if you have a very high resolution sensor and a narrow lens, the addition of an extender will not give any increase in resolution and may make it worse by inducing aberrations; if you have a low resolution sensor and a wide lens then the extender will tend to have its full effect. I've done loads of tests on the Rf 1.4x on the RF 100-500mm + R5 and find between about a 5-10% increase in resolution . With the 2xTC it can be up to 20% but for most of the time zilch - doubling up with Topaz can give better results. All history for me as the RF 200-800mm arrived today.

 

[docsmith]

You know, it is good to challenge assumptions as technology evolves. A RF 1.4xTC just arrived. I'll be testing it with the 200-800. But, I am now curious about the EF extenders with my 100-400 II and 500 f/4 II.

The theory makes sense, if you had infinite pixels and a perfectly sharp lens, you could crop forever and not need extenders. It is sounding like we might be approaching that in a few circumstances.

 

[AlanF]

You know, it is good to challenge assumptions as technology evolves. A RF 1.4xTC just arrived. I'll be testing it with the 200-800. But, I am now curious about the EF extenders with my 100-400 II and 500 f/4 II.

The theory makes sense, if you had infinite pixels and a perfectly sharp lens, you could crop forever and not need extenders. It is sounding like we might be approaching that in a few circumstances.

I haven't really tested the 1.4x with the RF 200-800mm yet as 800mm is I think going to be long enough for me in practice. Have you any interesting information?

 

[docsmith]

I haven't really tested the 1.4x with the RF 200-800mm yet as 800mm is I think going to be long enough for me in practice. Have you any interesting information?

Hi Alan,

Mostly impressions so far. AF suffers more than usual when I add a 1.4TC on a lens. Hit rate has suffered (this lens needs light). In terms of 1.4x improvement in resolution, probably a little, but likely not enough to justify the AF issues.

I am planning a comparison over the next few days. But here are a smattering of images at my feeder.

w/ 1.4xTC


For comparison, a Junco with the bare lens (without 1.4xTC) on the same branch at my feeder and butterfly bush next to my feeder:

 

[AlanF]

Hi Alan,

Mostly impressions so far. AF suffers more than usual when I add a 1.4TC on a lens. Hit rate has suffered (this lens needs light). In terms of 1.4x improvement in resolution, probably a little, but likely not enough to justify the AF issues.

I am planning a comparison over the next few days. But here are a smattering of images at my feeder.

w/ 1.4xTC
View attachment 214335View attachment 214336View attachment 214337View attachment 214338

For comparison, a Junco with the bare lens (without 1.4xTC) on the same branch at my feeder and butterfly bush next to my feeder:
View attachment 214343

View attachment 214342

View attachment 214341

I took a few shots of my charts with the 1.4x on the RF 200-800mm on the R5. No extra resolution over the bare lens, and possibly worse.

 

[docsmith]

I took a few shots of my charts with the 1.4x on the RF 200-800mm on the R5. No extra resolution over the bare lens, and possibly worse.

I haven't done my test yet, but I am not seeing any extra resolved details.

Also, and I believe you have referred to similar, but borrowing this from Optical Limits:



This is from their review of the 800 f/11. I am usually very impressed with what they do, so while I have not worked out their methods, I am willing to trust them to a large extent. Also, for the below discussion, OL estimated the 45 MP R5 sensor behaves more like a 34 MP sensor due to the Bayer Filter, etc.

Regarding the classic question, is the lens out resolving the sensor, is the sensor out resolving the lens and now we are looking at high enough aperture values, that we need to include the impacts of diffraction. Canon are not dummies. The 100-500L is likely out resolving the R5 sensor, and f/7.1 can go to f/10 and be right around 34 MP, so you get most of the benefit of the 1.4xTC on the R5 100-500 combo. The 200-800 may be designed to perfectly triangulate with the R5, the sensor and lens have similar resolving power, while 34 MPs at f/9 and f/10 being at or below the curve. But, push anything, and there is no gain. For example, 1.4x TC, 40% gain, but f/13 in the curve above is right around 20 MPs, which would offset the benefit of the tc.

Anyway, this figure adds to a conclusion I am coming too, that the RF 200-800 may well be best for cameras with pixel densities similar to the R5 and lower. This would include a good number of cameras including rumored R1, R3, R5, R6 I/II, and R8 with the R10/50/100 falling between the pixel density of the R5 and R7.

Just an idea I am playing with. Being an engineer...this is something I would do.

 

[AlanF]

I haven't done my test yet, but I am not seeing any extra resolved details.

Also, and I believe you have referred to similar, but borrowing this from Optical Limits:

View attachment 214358

This is from their review of the 800 f/11. I am usually very impressed with what they do, so while I have not worked out their methods, I am willing to trust them to a large extent. Also, for the below discussion, OL estimated the 45 MP R5 sensor behaves more like a 34 MP sensor due to the Bayer Filter, etc.

Regarding the classic question, is the lens out resolving the sensor, is the sensor out resolving the lens and now we are looking at high enough aperture values, that we need to include the impacts of diffraction. Canon are not dummies. The 100-500L is likely out resolving the R5 sensor, and f/7.1 can go to f/10 and be right around 34 MP, so you get most of the benefit of the 1.4xTC on the R5 100-500 combo. The 200-800 may be designed to perfectly triangulate with the R5, the sensor and lens have similar resolving power, while 34 MPs at f/9 and f/10 being at or below the curve. But, push anything, and there is no gain. For example, 1.4x TC, 40% gain, but f/13 in the curve above is right around 20 MPs, which would offset the benefit of the tc.

Anyway, this figure adds to a conclusion I am coming too, that the RF 200-800 may well be best for cameras with pixel densities similar to the R5 and lower. This would include a good number of cameras including rumored R1, R3, R5, R6 I/II, and R8 with the R10/50/100 falling between the pixel density of the R5 and R7.

Just an idea I am playing with. Being an engineer...this is something I would do.

I'll do a post about this later. I have recalculated the OL curve. It's misleading because it has just the effective megapixels from considering only diffraction and doesn't include that the overall effective number must be a combination of the actual physical pixels with the diffractive number. (They also use a shorter wavelength than normal, 420nm rather 500 or 550nm for green light, which I back calculated from their curve). The actual calculation is very complicated but I have derived a simplified solution that is not accurate but an improvement. Here is a quick plot of my calculations for 45 and 24 Mpx sensors at 500nm combining both sensor and diffraction limitations compared with the OL 420nm curve for just diffraction.

 

[docsmith]

Hi Alan,

I have been looking at the graph you generated and, while I am into this concept, I am wondering if it is a bit extreme. If I am reading it correctly, at ~f/11, which I shoot at frequently and produce very good/sharp results, my R5 would have less than 20 MP of resolution. I do not recall seeing much of an IQ drop off in my images going from f/4 and f/5.6 out to f/8-11. In the figure above, I should expect ~30% fewer equivalent MPs.

So, I took a second and tried to convert Optical Limits LW/PH values for the RF 28-70 f/2 (center at 40 mm) and convert those to equivalent MPs:

f/	OL Result (LW/PH)	R5 Sensor Vertical Pixels	%	Equiv MP
2	5119	5464	94%	39.3
2.8	5350	5464	98%	42.9
4	5192	5464	95%	40.4
5.6	4956	5464	91%	36.8
8	4665	5464	85%	32.6
11	4349	5464	80%	28.4

So, here I assumed "PH" is vertical (24 mm) of the sensor, so on the R5 5464 pixels. Then, I am also assuming that a row of effective pixels would be equal to a "line width".

I am happy to learn more about this, especially if I am off in my assumptions.
Thanks...

 

[AlanF]

Hi Alan,

I have been looking at the graph you generated and, while I am into this concept, I am wondering if it is a bit extreme. If I am reading it correctly, at ~f/11, which I shoot at frequently and produce very good/sharp results, my R5 would have less than 20 MP of resolution. I do not recall seeing much of an IQ drop off in my images going from f/4 and f/5.6 out to f/8-11. In the figure above, I shoudl expect ~30% fewer equivalent MPs.

So, I took a second and tried to convert Optical Limits LW/PH values for the RF 28-70 f/2 (center at 40 mm) and convert those to equivalent MPs:

f/	OL Result (LW/PH)	R5 Sensor Vertical Pixels	%	Equiv MP
2	5119	5464	94%	39.3
2.8	5350	5464	98%	42.9
4	5192	5464	95%	40.4
5.6	4956	5464	91%	36.8
8	4665	5464	85%	32.6
11	4349	5464	80%	28.4

So, here I assumed "PH" is vertical (24 mm) of the sensor, so on the R5 5464 pixels. Then, I am also assuming that a row of effective pixels woudl be equal to a "line width".

I am happy to learn more about this, especially if I am off in my assumptions.
Thanks...

Before seeing your post, I replaced my graph using a better equation that I used a couple of years ago and has a shape that fits measured mtf values from ePhotozine.

 

[AlanF]

Hi Alan,

I have been looking at the graph you generated and, while I am into this concept, I am wondering if it is a bit extreme. If I am reading it correctly, at ~f/11, which I shoot at frequently and produce very good/sharp results, my R5 would have less than 20 MP of resolution. I do not recall seeing much of an IQ drop off in my images going from f/4 and f/5.6 out to f/8-11. In the figure above, I shoudl expect ~30% fewer equivalent MPs.

So, I took a second and tried to convert Optical Limits LW/PH values for the RF 28-70 f/2 (center at 40 mm) and convert those to equivalent MPs:

f/	OL Result (LW/PH)	R5 Sensor Vertical Pixels	%	Equiv MP
2	5119	5464	94%	39.3
2.8	5350	5464	98%	42.9
4	5192	5464	95%	40.4
5.6	4956	5464	91%	36.8
8	4665	5464	85%	32.6
11	4349	5464	80%	28.4

So, here I assumed "PH" is vertical (24 mm) of the sensor, so on the R5 5464 pixels. Then, I am also assuming that a row of effective pixels woudl be equal to a "line width".

I am happy to learn more about this, especially if I am off in my assumptions.
Thanks...

I did a whole series 4 years ago of these plots for the 5DSR using highest values from the best lenses published by ePhotozine and Photozone (aka opticallimits) as there was a lot of data available. I've dug out the old plots. The MTF50s at f/11 drop from the theoretical maximum of 5792 by about 30%. This is equivalent to the 50 Mpx sensor dropping to about an effective 25 Mpx at f/11. Using my equation (which is an approximation and ignores filters etc) and wavelength of 420 nm, it predicts the R5 is roughly equivalent to a 23 Mpx sensor and at 500nm 17.6 Mpx. Take those as ballpark figures.

 

[AlanF]

To my then surprise, Lenstips charts on the 5DIII had a similar drop at f/11, despite the much lower Mpx.

 

[docsmith]

So, I am still working to wrap my head around the implications of some of this. First, and the point of this, the fall off as you go from f/9 to f/13 can be substantial and may be enough to negate any benefit of a TC on the 200-800. Ok, I am seeing that.

But, some of those slopes are nearly linear, implying that a sensor is at peak resolution with large aperture lenses. This flies in the face of the conventional wisdom that sharpness at most lenses peaks around f/5.6-f/8, then diffraction starts taking its toll. Now, sure, there could be two elements in play, Lens issues impacting wide-open shots and the diffraction affecting larger F number images with f/5.6-8 being the "sweet spot"....sure...except, the data doesn't really show that.

So, I do mostly enjoy photography for the sake of actual photography. But I'll enjoy trying to understand the science/mechanics behind it better. Helps understand how remarkable it all really is. So, I see some tests in my future.

BTW, cranking the Lenstip numbers for the 5DIII, ~52 lp/mm is equivalent to about a 9 MP sensor. Ouch. But, I remember looking at Lenstips and their general guide, >40 is good, >45 extremely good, and as I recall >30 was passible. ....30 lp/mm is ~3 MP (30 lp/mm x 2 pixels/lp x 36 or 24 mm). Yikes!

 

[AlanF]

So, I am still working to wrap my head around the implications of some of this. First, and the point of this, the fall off as you go from f/9 to f/13 can be substantial and may be enough to negate any benefit of a TC on the 200-800. Ok, I am seeing that.

But, some of those slops are nearly linear, implying that a sensor is at peak resolution with large aperture lenses. This flies in the face of the conventional wisdom that sharpness at most lenses peaks around f/5.6-f/8, then diffraction starts taking its toll. Now, sure, there could be two elements in play, Lens issues impacting wide-open shots and the diffraction affecting larger F number images with f/5.6-8 being the "sweet spot"....sure...except, the data doesn't really show that.

So, I do mostly enjoy photography for the sake of actual photography. But I'll enjoy trying to understand the science/mechanics behind it better. Helps understand how remarkable it all really is. So, I see some tests in my future.

BTW, cranking the Lenstip numbers for the 5DIII, ~52 lp/mm is equivalent to about a 9 MP sensor. Ouch. But, I remember looking at Lenstips and their general guide, >40 is good, >45 extremely good, and as I recall >30 was passible. ....30 lp/mm is ~3 MP (30 lp/mm x 2 pixels/lp x 36 or 24 mm). Yikes!

Same here - love photography and the science behind it, which helps me get the best from my gear. The conventional wisdom is out of date in many cases. Modern lenses are often much sharper wide open than in the past.

 

[docsmith]

Hmmm...posting results with a RF 200-800 in a thread about the RF 100-500?

It is where we are talking about aperture's impact on IQ....so, continuing this conversation, here are results at different apertures of Juncos at my birdfeeder. Severe crops once again.

R5, RF 200-800 @ 707 mm, ISO 1600

1/1600, f/9


1/800, f/13



1/400, f/18




Adding the 1.4xTC

990 mm, 1/800, f/13


990 mm, 1/400, f/18



I am seeing some degradation with increasing aperture, but I would not call it severe. Mostly a loss of contrast, which makes sense if you think of MTFs being white/black blocks and image degradation is really just the blurring of those blocks. But, it actually renders "ok" IMO.

All the above images had no processing. This (990 mm, cropped, 1/400, f/18, ISO 1600) is very quickly processed in LR, run through denoise, 50 sharpening (+10), +5 texture, etc.



So, does aperture affect IQ in this example, a little, but it doesn't fall apart, IMO.

Does the 1.4x TC add to the IQ? I am not seeing significant extra detail. In fact, after staring at several of these side by side, I preferred the 707 mm crop better than the 990 mm crop. Similar detail and better contrast. I'll test on my 500 f/4, but I can see the 1.4xTC being returned at this point.

 

[AlanF]

Hmmm...posting results with a RF 200-800 in a thread about the RF 100-500?

It is where we are talking about aperture's impact on IQ....so, continuing this conversation, here are results at different apertures of Juncos at my birdfeeder. Severe crops once again.

R5, RF 200-800 @ 707 mm, ISO 1600

1/1600, f/9
View attachment 214486

1/800, f/13
View attachment 214493


1/400, f/18
View attachment 214488
View attachment 214489


Adding the 1.4xTC

990 mm, 1/800, f/13
View attachment 214490

990 mm, 1/400, f/18
View attachment 214491


I am seeing some degradation with increasing aperture, but I would not call it severe. Mostly a loss of contrast, which makes sense if you think of MTFs being white/black blocks and image degradation is really just the blurring of those blocks. But, it actually renders "ok" IMO.

All the above images had no processing. This (990 mm, cropped, 1/400, f/18, ISO 1600) is very quickly processed in LR, run through denoise, 50 sharpening (+10), +5 texture, etc.
View attachment 214492


So, does aperture affect IQ in this example, a little, but it doesn't fall apart, IMO.

Does the 1.4x TC add to the IQ? I am not seeing significant extra detail. In fact, after staring at several of these side by side, I preferred the 707 mm crop better than the 990 mm crop. Similar detail and better contrast. I'll test on my 500 f/4, but I can see the 1.4xTC being returned at this point.

I think the reason why you are seeing little decrease in resolution with increasing f-number is because the feather/barb spacing on the Junco is basically equivalent to relatively widely separated line pairs compared with the pixels on the sensor. The spacing when I stare at enlarged images looks like they make line pairs spanning 4-8 pixels. As the diffraction blurring increases, it will first manifest itself as a loss of contrast and when the contrast drops to about 0.1 they will merge. When I look at my charts I can see black/white line pairs resolved down to the Nyquist frequency of 1 lp per 2 pixels for a sharp lens even at the diffraction limit, like with the 100-500mm at f/7.1.

Following on from this, You won't see increased resolution with a 1.4x TC if the effective line spacing of the detail in your image is 2-3x lower than the resolution of you sensor. It's only if the details are at the limit of resolution and the extra 40% of length brings details from just below resolution to just above will you see the benefits. I've just done a day or two of taking images from 500-800mm with the 200-800mm, and some will be shocked to find that the extra focal length makes very little difference for much of the time. I'll post a thread on this.

 

[docsmith]

I've just done a day or two of taking images from 500-800mm with the 200-800mm, and some will be shocked to find that the extra focal length makes very little difference for much of the time. I'll post a thread on this.

I am seeing the same thing. I tested at 600/700/800 mm. Cropping down, it is remarkably similar. All very good, mind you. I do not want to get lost in details, in my previous test, only my 500 f/4 II w 1.4tc beat this lens. But it is very interesting that the extra zoom length is not necessarily providing extra details on the R5.