How much better is MedFormat digital?

What was asked for...

bobbyz wrote in post #17330980
Can we get some sample shots from canon and then with this fiji to show the sony like improvements.

2nd, can we see some medium format shots comparing to Nikon/Sony sensors?.

I find two different shots from two cameras totally useless for 'comparison' purposes. The same scene shot from the same tripod, one immediately after the other, has better comparison value.

Wilt wrote in post #17346471
What was asked for...

I find two different shots from two cameras totally useless for 'comparison' purposes. The same scene shot from the same tripod, one immediately after the other, has better comparison value.

I completely agree. Random shots compared tell you almost nothing.

I should be clear here about 35mm digital vs. digital medium format:

1. In colors, both are very similar or closer, but MF has an edge, it could be that bigger sensor or better DR or more bit, is it 16bit vs. 12/14bit of 35mm?

2. Resolution, i mean pixel, i think here we can't talk about comparison, the largest mp 35mm camera are Sony/Nikon 36mp, i started digital medium format with 39mp and that was in 2009 if i remember, now i have 60mp, no comparison

3. Exposure DR, honestly speaking, MF has that since long time and most recent 35mm are coming closer, but if i we have to choose 1 winner then which best 35mm is on top than the best digital medium format? don't compare worse old medium format with latest high mp better DR, this is not comparison, this is just misleading, my best Canon and even my Sony A7R don't come close for highlight/shadow recovery against my H4D-60, the only thing with Canon and Sony are better is only cleaner recovery that's all.

4. Tonality or rendition, if we talk about colors then i said above about color, if about sharpness, mostly lenses are the key here, but if we put lens performance with sensor size and whatever another things, this must have a speak, even my normal lens of digital medium format which is HC 80mm giving me sharpness and tonal of skin color than not my best sharpest Canon Sony lens can produce, it come closer yes as i said, but never same or out-resolve, so if i have the choice between digital medium format and 35mm digital to shoot portraits under the light then i will always choose medium format even i can do amazing great with top line 35mm digital.


So, if we come to 35mm digital, the following are the mainly factors that give them advantages over digital medium format, not sure for how long but no doubt for long enough maybe:

1. Price, this is the first factor and i am sure if digital medium formats i mean brand news were around $10k or less, then i will see many medium formats in gear list of members online in many photography websites, and i will not see many jealous comments, even in 35mm that happen between expensive and affordable 35mm DSLR or mirrorless gear.

2. Weight, this is an issue for many shooters, all or most medium formats are bulky and big sizes, and many don't like this as daily use or travel or all-around shooting, so understandable.

3. AF speed, i am sure if that medium format has this ability against 35mm then we will see more shooter for MF, but i think even fastest AF MF is still way way beyond best AF 35mm DSLRs such as 1DX or 5D3 and so on.

4. High ISO, i can't remember one topic here or there when talking about newer 35mm digital without talking and showing about high ISO performance, even it makes me thinking as it is the only reason to buy the latest body due to best high ISO capability, but don't take me wrong, it is not the only reason, it is just my feeling about it.

5. FPS, for many shooters of sports, wildlife/ birds and so this is very handy, i know many will talk about timing, well, i used different cameras and i always get more keepers with higher fps cameras than slower one even both can be in same AF league, this is feature to use not to ignore it, if someone is a big master in timing then i hope to see all those SI or ESPN using film until now, they all used film in the past, why not keep using it, the question is, why they use top line of Canon/Nikon to shoot if they can do great with entry-level? i didn't see 7D or 7DII shooters anywhere in those pro sports shooter, even if they do i can guess it is 1% against 1 series or Nikon Dx gear.

6. Cocktail of lenses and accessories, yes, in medium format the most lenses for one system i can see is around 15 glasses if we talk about only manufacturer lenses, but if we talk about 3rd party or another brands to be used then even 35mm digital will lead here, if i can use 30 lenses with medium format then i can use over 100 for my 35mm digital, any objection here?


At the end, this is my opinion above about 35mm digital against digiMF, don't take my post as an accurate answers, you can correct me please or add, just don't joke much because i have wrong things, we all learn and we all have opinions, i may love MF a lot but many others may hate or disagree with that, so no need to make war about different points or opinions, if you have something to add or correct then let's do that politely, and when i say that it is because when i just recently replied[POSTed] in FB about something photographic, someone posted to show how wrong we are and he defended his points in so much way that we shouldn't talk again.

Comparison just for comparison can be done, but what is the purpose of it if it will not go anywhere? many agree about MF quality but still they can't buy it or use or afford it, then why ask for it, i can put my H4D-60 in test against all my Canon/sony gear, but at the end, what you gonna do with that or help you in what?

Sorry for my way of posting

Very nice post Tarek. I am new to medium format but I can already confirm everything you said. I will try to make comparisons (same shots) with the H5D, 5D3 and A7R.

RedCatPhoto wrote in post #17349943
Very nice post Tarek. I am new to medium format but I can already confirm everything you said. I will try to make comparisons (same shots) with the H5D, 5D3 and A7R.

Cool Jean, do that, and i am sure your results will be better than what i can do, i have better but older MF than yours, but i don't have better Canon DSLR than yours in mp [5D3 vs. 1DX], and we are equal in A7R, so better you make that comparisons than mine.

CRCchemist wrote in post #17323670
This is what I have long wondered, but it had been disputed by many people.

There is a technique using a program called FIJI to apply a Fast Fourier Transform to a Canon sensor image, that virtually eliminates all pattern noise -- down to the level of the Sony and Nikon cameras. That allows you you to extract the same amount of dynamic range and detail that the Sony and Nikon cameras have from an image taken by a Canon camera that is suffering from noise at low ISO and has been boosted several stops in development software.

Anyway, I'm just confirming what you said. The advantage of Sony doesn't appear to be intrinsic to the sensor, it seems to be due to the algorithms that Sony programs into their chips that manipulate the image before it is saved as a RAW file on the memory card.

I'd love this to be true, but I am a bit dubious, as it flies in the face of much evidence regarding the architectural differences between Canon's sensors and (most specifically) Sony's Exmor.

As far as I'm aware, it would be hard to patent noise removal via FFT; such that other companies couldn't stop Canon doing it... so if it's as "simple" as an additional processing step then why haven't Canon done it already?

It also occurs to me that there's going to be a difference between noise removal and, assuming the sensor architecture on the Exmor is truly different, not adding the noise in the first place.

Until I saw controlled testing of D810/a7R shots vs 5D3/6D (plus noise reduction) I'd remain sceptical; even though it would be great if it were true.

sploo wrote in post #17350307
I'd love this to be true, but I am a bit dubious, as it flies in the face of much evidence regarding the architectural differences between Canon's sensors and (most specifically) Sony's Exmor.

As far as I'm aware, it would be hard to patent noise removal via FFT; such that other companies couldn't stop Canon doing it... so if it's as "simple" as an additional processing step then why haven't Canon done it already?

It also occurs to me that there's going to be a difference between noise removal and, assuming the sensor architecture on the Exmor is truly different, not adding the noise in the first place.

Until I saw controlled testing of D810/a7R shots vs 5D3/6D (plus noise reduction) I'd remain sceptical; even though it would be great if it were true.

If you were to completely eliminate the pattern noise you'd bring the 5D3 into the realm of the 7D2, but that still wouldn't get you the dynamic range of the Exmor chips since the pattern noise is not the limiting factor. You can, however, use MagicLantern to get dynamic range equivalent to Exmor out of the 5D3/6D sensors.

raptor3x wrote in post #17351388
If you were to completely eliminate the pattern noise you'd bring the 5D3 into the realm of the 7D2, but that still wouldn't get you the dynamic range of the Exmor chips since the pattern noise is not the limiting factor. You can, however, use MagicLantern to get dynamic range equivalent to Exmor out of the 5D3/6D sensors.

I assume there's "noise" and there's "pattern noise". I'm told the the 6D is relatively free of banding, so it wouldn't surprise me if Canon have worked out a way of removing/hiding the most objectionable (noticeable) problem. However, it doesn't mean the shadow noise has overall been reduced to Exmor levels, so that would explain a continued deficit. I'm speculating there: I haven't played with a 6D or 7D2.

The magic lantern dual iso trick is very clever, but the resolution loss takes away some of the practicality for me (as well as needing to run ML, which last time I looked wasn't trivial on a 5D3).

Wilt wrote in post #17316032
The reader needs to keep in mind that optics is a very important factor in resolution...one CANNOT assume that a 40% pixel resolution in one direction will necessarily result in 40% better detail resolution. For example, if we double the pixel count on APS-C cameras, from 8Mpixel (350D) to 15Mpixel (50D), we see detail rise only 34% (2640 vs. 1835 LP/PW) with the 70-200mm f/2.8 lens...

It used to be accepted that, as film sizes went up, lens resolution went down. Lots of lines/mm resolution just wasn't necessary, especially for 8" X 10" contact prints.
If I remember right, very good for 35mm was 110 lines/mm. For 2-1/4" it was about 80 lines/mm.

sploo wrote in post #17354142
I assume there's "noise" and there's "pattern noise". I'm told the the 6D is relatively free of banding, so it wouldn't surprise me if Canon have worked out a way of removing/hiding the most objectionable (noticeable) problem. However, it doesn't mean the shadow noise has overall been reduced to Exmor levels, so that would explain a continued deficit. I'm speculating there: I haven't played with a 6D or 7D2.

The magic lantern dual iso trick is very clever, but the resolution loss takes away some of the practicality for me (as well as needing to run ML, which last time I looked wasn't trivial on a 5D3).

You've got the right idea, there's the random read noise and then the pattern noise. The 7D2, and to a lesser extent, the 6D have eliminated most of the pattern noise but the random read noise is still there. The DxO numbers that show the 2-3 stop advantage are really only looking at the random noise so even with the pattern noise gone you're still pretty far off from the Exmor sensors.

The resolution loss from DualISO is almost trivial in the later builds although it was fairly significant in the first few releases. Installing ML on the 5D3 is very simple now and the only real downside to having it installed is the increased startup time as it loads the ML modules, but that can be mitigated as well.

raptor3x wrote in post #17359433
You've got the right idea, there's the random read noise and then the pattern noise. The 7D2, and to a lesser extent, the 6D have eliminated most of the pattern noise but the random read noise is still there. The DxO numbers that show the 2-3 stop advantage are really only looking at the random noise so even with the pattern noise gone you're still pretty far off from the Exmor sensors.

The resolution loss from DualISO is almost trivial in the later builds although it was fairly significant in the first few releases. Installing ML on the 5D3 is very simple now and the only real downside to having it installed is the increased startup time as it loads the ML modules, but that can be mitigated as well.

Interesting post; thanks.

It does explain the perceived improvement in the newer Canon bodies at least.

ML was permanently modifying something in the 5D3 when installed (last time I looked), so that put me off a bit. I really ought to install it on my old 7D to try the features.

sploo wrote in post #17359455
Interesting post; thanks.

It does explain the perceived improvement in the newer Canon bodies at least.

ML was permanently modifying something in the 5D3 when installed (last time I looked), so that put me off a bit. I really ought to install it on my old 7D to try the features.

Ah, they fixed the bootflag issue with the latest big update to ML. It's very simple now to completely eliminate ML from the 5D3.

Sorry to post...after a yr. But this Fiji software the "chemist" posted about got me intrigued.

So I tried it.
Original file, then the mask.

The mask and then The result

That couldnt be right...Looks much worse and it just lightened it. WRONG.
So I inverted the mask/filter. and Sucess. Before and after.

Looks like a viable thing, but...Just shoot at the right ISO, which is 100. hehe.
Info for those wondering. Canon 5Dc, iso 400, pushed 3 stops in ACR, so 3200 effective.
The OP wasnt clear about getting rid of color noise, which I had, so I did +50 in ACR process 2003,
not the new 2012, which makes it a new camera, almost..