Why does high t stop start to get soft ?

[PatrickFaith]

I had a problem today, where i got lazy and didn't put in a ND filter on and just cranked the T stop past T20 on my coverage/2nd camera doing wide shots. I normally don't shoot at a high T stop, and the shot was a bit soft and unusable for vfx matte work. Back of my brain was thinking the higher the T stop, the more clarity the lens would have, since less of the lens is used. But after this little screw up (i fortunately really didn't need the 2nd camera shot), I did a bit of testing and found I often don't like the lens above T16. I was thinking maybe at high t-stops you start getting interference across the aperture or some other "physics" property?

 

[Ian Tetzner]

Diffraction: http://www.cambridgeincolour.com/tutorials/diffraction-photography.htm

 

[Bob Gundu]

Higher T stops makes more in focus but doesn't necessarily means sharper. Most lenses are optimal at T5.6 - T8.

 

[Zeb B]

Most lenses are sharpest at 3-4 stops above wide open. Basically, the middle stops usually T8-11

 

[PatrickFaith]

Diffraction: http://www.cambridgeincolour.com/tutorials/diffraction-photography.htm

Excellent website, even lists different camera's effect the f-stop. Hmmm ... for dragon, since the pixels got smaller, that means the highest usable T stop got lower.

Higher T stops makes more in focus but doesn't necessarily means sharper. Most lenses are optimal at T5.6 - T8.

So for the dragon, I guess the diffraction start's getting larger then the pixel at f8'ish, at f16 one pixel then smears across 9 pixels (i.e. shooting at 6k is no different then shooting at 2k).
So my rule of thumb is never go above T8 now ? (my eye notices it even on a ok monitor at 16, t22 it looks bad, so on a 4k monitor when i look at it in post ... i never should go above T12).

btw on 1080p/2k camera's i didn't notice this at all (because I assume either the pixels are larger or they are summing them so the "blur" doesn't matter).

 

[PatrickFaith]

Most lenses are sharpest at 3-4 stops above wide open. Basically, the middle stops usually T8-11

The dragon and the master primes are the best combination i've ever had, I'm finding I really need to up my game in order to use them effectively.

The master prime anamorphics with dragon sensor size I think are solid between T2 and T8, then start to degrade from there based on diffraction (i.e. the lens quality i think is great all the way to T22). It's apparent on 720p monitors above T16, looks bad above T20 on 1080p. Most of the time lens meta data isn't getting to me very often that shows T stop, so I wasn't sure why the picture looked a bit fuzzy when I was doing mattes & reframing - now i know.

I was pondering on getting some medium format lenses that tended to start with higher f-stops, makes me rethink some things as I don't want to go over f8 due to this basic physics problem Ian referenced. I btw don't remember anyone on reduser stating that the dragon sensor really limits lenses to f8/t9 to maintain full resolution(this is basic physics, nothing per se to do with RED ... but I bet a lot of other people are like me and didn't get this diffraction limit equation which is basically related to the f-stop by single sensor/pixel size.

(edit ... did find a link that somewhat went over this ... of course Phil explained it a bit ... http://www.reduser.net/forum/showthread.php?131746-Red-vs-Alexa-pixel-dimensions, when i work out the equations basically red dragon is good to f8 while Arri Alexa can go to f11).

 

[Phil Holland]

Heh. Everybody posted the good info tonight. I ate my tacos without replying

Typically you'll be looking for the sweet spot on most optics around 2-3 stops stopped down.

This is one of the reasons that the mysterious 2.8/4 split is so commonly used on most super and standard speed lenses.

I remember my first ever shoot with Zeiss Master Primes forever ago, which we used instead of Ultras (which are already stellar), we shot at T4 and I damn near cut myself on the images.

Zeb is right by going up to 4 stops as well. There is a range. If I recall correctly around T8-11 you see diffraction creep in on Summilux-C and Master Primes on Mysterium-X 5K. Dragon is a bit different.

 

[David Battistella]

On my Leica R set I like to ride them at about 8/5.6 on Dragon. Some stuff I will go to 11 (right landscapes, etc. I just did some stuff I rated at ISO 1000 just to get to this stop and the results on dragon are quite pleasant.

Great thread, with great advice!

We need many mare threads like this one.

 

[Michael Tiemann]

When it comes to DRAGON Monochrome, everything said about f8+ and 0.5 micron pixel pitch is consistent.

When it comes to color, the DRAGON bayer pattern uses at least 4 sensor pixels per image pixel. The RED OLPFs smear a point source of light (actually, an airy disc) to cover those 4 pixels (or more?) to prevent moire. There are already 3rd-party OLPFs with interesting properties. Perhaps there's a market for a DOF OLPF, one which assumes that there's at least a 1 micron diffraction smear. That would give sharper images for when higher depths of field were desirable, making f16 work as well as f8 with a conventional OLPF.

Note that I'm using f-nomenclature instead of T-nomenclature because transmission is irrelevant to the calculation of the airy disc. It's all about the f-stop ratio.

 

[Michael Lindsay]

Rule of thumb that I use re diffraction

On wides you can just about see diffraction one to two stops down from the reciperical of the pixel pitch... on CU stuff you can push 2+ stops further...

 

[PatrickFaith]

Good news is I decided to reshoot yesterdays footage with my new insight ... lmao I had to reshoot a scene where this snake had just crawled through some grass, and while i was reshooting the camera moved over a real snake skin that just happened to be there. Completely made up for the reshoot.

Heh. Everybody posted the good info tonight. I ate my tacos without replying
Typically you'll be looking for the sweet spot on most optics around 2-3 stops stopped down.
This is one of the reasons that the mysterious 2.8/4 split is so commonly used on most super and standard speed lenses.
I remember my first ever shoot with Zeiss Master Primes forever ago, which we used instead of Ultras (which are already stellar), we shot at T4 and I damn near cut myself on the images.
Zeb is right by going up to 4 stops as well. There is a range. If I recall correctly around T8-11 you see diffraction creep in on Summilux-C and Master Primes on Mysterium-X 5K. Dragon is a bit different.

Thanks Phil, I hope I didn't interrupt you eating your vegetables though

I think what I'm going to do with master primes is just follow your advice and keep them under T8, especially if I will be using a frame from a motion sequence for vfx. If it's a shot I don't need to reframe or use for matte's, i'll keep to below T11.

On my Leica R set I like to ride them at about 8/5.6 on Dragon. Some stuff I will go to 11 (right landscapes, etc. I just did some stuff I rated at ISO 1000 just to get to this stop and the results on dragon are quite pleasant.

Great thread, with great advice!

We need many mare threads like this one.

Agreed David, especially since I'm off shooting solo most of the time I find when I am doing something totally wrong, it often isn't caught. I'm not sure exactly what 8/5.6 means ... i noticed Phil also used a 2.8/4

When it comes to DRAGON Monochrome, everything said about f8+ and 0.5 micron pixel pitch is consistent.

When it comes to color, the DRAGON bayer pattern uses at least 4 sensor pixels per image pixel. The RED OLPFs smear a point source of light (actually, an airy disc) to cover those 4 pixels (or more?) to prevent moire. There are already 3rd-party OLPFs with interesting properties. Perhaps there's a market for a DOF OLPF, one which assumes that there's at least a 1 micron diffraction smear. That would give sharper images for when higher depths of field were desirable, making f16 work as well as f8 with a conventional OLPF.

Note that I'm using f-nomenclature instead of T-nomenclature because transmission is irrelevant to the calculation of the airy disc. It's all about the f-stop ratio.

I noticed this and it was driving me crazy, I wasn't correlating it to my obvious mess up with using too high of a t-stop ... and started going down a total bogus olpf issue with back focus ... when the simple answer is pay attention to my t-stop. When I was shooting with a monochrome epic, and didn't really notice this nearly as much.

But when i moved to color with the debayer pattern and also the smaller sensor, also started using a better matte box so had greater contrast, and completely moved from my nikon lenses to arri master prime anamorphics - all these together made this problem just "popped" out at me. I think full cinema rigs are a bit like a formula one car, I was a newb and the first thing i did was drive my camera right into a wall.

Rule of thumb that I use re diffraction

On wides you can just about see diffraction one to two stops down from the reciperical of the pixel pitch... on CU stuff you can push 2+ stops further...

That's a great rule, so just to double check I understand the rule ... if the pitch is 5 (i.e. dragon) then start to worry when i got above f7 (i.e. t8).

 

[Phil Holland]

Just to add to this discussion. I typically work between wide open and T16. I don't recall many times in my life shooting more stopped down than that for motion picture production. Most productions I've shot or worked on float in that T2.8/4 split or a bit higher or lower. Some of the brave shoot wide open, which I do a lot as well, but it truly depends on the lenses and piece as to what I'm doing exactly.

There is one additional thing to think about when managing diffraction. That's oversampling and scaling down to your finished resolution.

Actually one other additional thing. I've seen moire and aliasing on several cameras in my life. Diffraction can help soften up an image just enough where that might vanish off the monitor. Haven't had to do that often, but if you are shooting skylines or cityscapes it's a handy thing to know with certain cameras.

 

[Antony Newman]

Is there a mode where 4 times the frame rate is used @ 360 degrees, but only two consecutive shots are stored?
ie frames 1,2, 5,6, 9,10, ...

This might 'rescue' a f16 shot - allowing an f8 shot.

+) one more stop DR at the expense of twice the data recorded.
+) or if the averaging was done in the camera before storing the data (but this might require some Weaponised grunt?)

Not sure if the read / reset of two combined frames would be noticeable?

AJ

 

[Cliff Hokanson]

Check your back focus

 

[PatrickFaith]

Is there a mode where 4 times the frame rate is used @ 360 degrees, but only two consecutive shots are stored?
ie frames 1,2, 5,6, 9,10, ...

This might 'rescue' a f16 shot - allowing an f8 shot.

+) one more stop DR at the expense of twice the data recorded.
+) or if the averaging was done in the camera before storing the data (but this might require some Weaponised grunt?)

Not sure if the read / reset of two combined frames would be noticeable?

AJ

Good idea, I use to do something like this to get extremely high resolution. As a side note, one thing I noticed when shooting in the sun, is the camera fan was cranking all the time and I could feel a slight vibration in my tripod. I btw do have a bunch of contact mics, so with that frequency and frame manipulations I bet I could get a real nice shot out of something that was a bit fuzzy ... if I put some mental energy into it. I have no idea at what level vibration on a tripod (from passing cars, wood floors, fans, etc ... ) is noticeable at the sensor size level of the dragon. In the ancient days with mechanical things going I use to put a weighted sand/lead shot bag under the camera on the tripod to dampen things.

Check your back focus

30 years ago I did a lot of technical optical work, checking for collimation and mounting problems(I pretty much forgot everything I use to do, plus I was following instructions "exactly" so didn't really understand what I was doing). The nice thing about renting is this type of thing was checked (also really in the old days very few people were even allowed to touch the camera), but in the vdslr scene I basically ignored this. I use to do this a bit with changing mounts on my cp2, but I need to get a process to really check out my lenses at least once a month. My old tools I use to use were super expensive, and I had a nice optical bench.

I am pondering putting together a bench for weekly camera testing/validaton (low budget, but at least a consistent space), any examples of what people are using for basic lens testing and calibration I would be very appreciative on. I have found a few articles ... http://www.theblackandblue.com/2011/07/27/checking-backfocus/ , but I need some exact procedures and products for testing master prime anamorphics to dragon sensors. I also really need to order a bunch of test charts, and figure out nice ways of mounting them without getting wrinkles, then putting that in my bench area. Also any devices to put on the mount to speed this process up that people use, would be most helpful.

 

[Maik Müller]

Is there a mode where 4 times the frame rate is used @ 360 degrees, but only two consecutive shots are stored?
ie frames 1,2, 5,6, 9,10, ...

This might 'rescue' a f16 shot - allowing an f8 shot.

+) one more stop DR at the expense of twice the data recorded.
+) or if the averaging was done in the camera before storing the data (but this might require some Weaponised grunt?)

You want to play with "Frame Summing" and "Frame Averaging" under settings / recording / frame processing.

 

[Brian Merlen]

recently instead of diffusion filters i shot something with no ND at t16-20ish and its not always a horrible thing i think, particularly if u can't light the place for lots of ND...

 

[David Mullen ASC]

My general belief is that you can stop down more with zooms, maybe because they aren't that fast and are telephoto lenses at the long end... But then, usually if I'm at the 290mm end of an Optimo at f/16, I'm shooting close-ups, not wide shots...

 

[Scot Yount]

This thread is fantastic, I learn so much here. I recently started using a RED Pro Zoom 18-85 (my first cinema lens) and I was out shooting some sailboats from a friend of mine's house, right on the Atlantic Ocean. It was super bright out and so, to keep my histogram in check at the lowest ISO my Scarlet-MX can do, i needed to go all the way to T22...at least that's what I thought. You can imagine, the pictures I brought back looked horrible across all focal lengths. I now know that the correct amount of ND would get my lens into the proper T stop, but I didn't have any with me at the time, and as stated, frankly didn't know any better. After talking to a still shooter friend of mine, I reshot the scenes using very low shutter angles and got pretty good results. My question is, on motion no faster than a sail boat, what are my disadvantages in cutting shutter angle down, or on faster motion for that matter?

 

[Michael Tiemann]

My question is, on motion no faster than a sail boat, what are my disadvantages in cutting shutter angle down, or on faster motion for that matter?

That's a whole 'nother can of worms...

In this RED tutorial you can clearly see that a 45 degree shutter introduces a kind of stroboscopic effect: http://www.red.com/learn/red-101/shutter-angle-tutorial

I suspect that as project frame rate increases, the sensitivity to motion blur may actually decrease (i.e., if the project is at 60 fps, it may be that there's so little relative motion blur at 1/120th of a second that 1/250th or 1/500th of a second doesn't make a real difference perceptually). But at 24fps (or there abouts) there's clearly lots of motion blur working for you at 1/48th of a second, and you will notice as you cut that down.

Don't neglect how much blur panning needs to look smooth.