Topics

Depth of Field question

Argyris_Theos_cml
 

A question addressing lens experts:
 
Suppose we have two cameras: one carrying S35 8K sensor, the other is FF35 with an HD sensor
I  employ a ~50mm lens on the first, a ~75 on the other. (I choose the exact focal lengths to match field of view)
I place my object at 3 meters and want my Depth of Field to match in both cameras, holding focus from 2.5 to 4 meters. 
The shot wiill be screened in a 10 meter wide screen, via a digital 8K projector. 
The only  post production work allowed is minimal color correction and uprezing of the HD image to 8k at the final stage, in order to avoid using the scaling machine of the projector server. 
Question:
At what f stop do I need to shoot on each camera to match DoF ?
 
Best Regards

Argyris Theos, gsc
DoP Athens Greece
cml@...
skype Argyris Theos

Mike Nagel
 

if my math doesn’t betray me…

 

~ T10.5 on S35

~ T15 on FF

 

 

 

- Mike Nagel

Director/Producer

L.A.

_._,_._,_

Argyris_Theos_cml
 

Mike,
What are the circles of confusion you are using ?
How did you choose them?
A. 

Argyris Theos, gsc 
DoP, Athens Greece,
+306944725315
Skype Argyris.Theos
via iPhone

31 Μαρ 2018, 12:41 μ.μ., ο/η "Michael Nagel via Cml.News" <michanagel=yahoo.com@...> έγραψε:

if my math doesn’t betray me…

 

~ T10.5 on S35

~ T15 on FF

 

 

 

- Mike Nagel

Director/Producer

L.A.

Mike Nagel
 

0.03 mm for both

 

What CoC are you using ?

 

 

- Mike Nagel

Director/Producer

L.A.

Geoff Boyle
 

In the lens comparisons I used 2 stop's between FF & S35
Whilst not exactly right it works well with the built-in ND's
Oh and posting which lens you prefer here will have no effect on the outcome which has been gradually changing.
Still need more votes otherwise the only way to find out which is which is to ask Carey at NAB as he's the only other person who knows.

Cheers
Geoff Boyle NSC
Cinematographer
EU based
+31 (0)637155076

Argyris_Theos_cml
 

The point is that you need different CoC because you have a different native resolution. 

Argyris Theos, gsc 
DoP, Athens Greece,
+306944725315
Skype Argyris.Theos
via iPhone

31 Μαρ 2018, 1:16 μ.μ., ο/η "Michael Nagel via Cml.News" <michanagel=yahoo.com@...> έγραψε:

0.03 mm for both

 

What CoC are you using ?

 

 

- Mike Nagel

Director/Producer

L.A.

Keith Putnam
 

Steve Yedlin, ASC addresses the issue of matching FOV and DOF (more precisely, "blur circles") on different format sizes in this document: http://www.yedlin.net/170504.html
It's very straightforward and a fascinating read. There's a lengthy intro which may feel redundant to you; he gets into the meat of the issue about halfway through.

Keith Putnam
Local 600 DIT
New York City

Mike Nagel
 

CoC calculation, AFAIK, has to do with size of image plane diagonal, not with resolution. For example, film has no “resolution”.

 

In any case, I was unclear before:

 

The different image plane is accommodated via crop factor. The 35mm FF reference CoC I used was 0.03.

 

I used a 1.6 crop factor for S35. You did not specify whether it’s S35 3 perf or 4perf target area.

 

Another slight deviation: DOF plane is 2.42m – 3.93m – close to what u wanted.

 

I did this quickly on the fly, u can adjust as needed... or maybe my math was off... what is your result ?

 

 

Thanks.

 

 - Mike Nagel

Director/Producer

L.A.

Argyris_Theos_cml
 

Excuse but you are totally wrong in saying :
"CoC calculation, AFAIK, has to do with size of image plane diagonal, not with resolution."

CoC is defined by magnification.
That's why I specified Screen Size
The CoC is the smallest projected circle that cannot distinguished (by naked eye) from a single point.
So we begin with a single recorded point (and this is not synonymous to a single pixel) but on the smaller sensor, in my hypothesis, we get 4 times the pixel count of the bigger sensor. That is per dimension, aka 16 times in total.
In such a case I would think that the CoC for this case's S35 sensor might be significantly smaller than this case's FF35 sensor.
Allow me to repeat that the S35 sensor would be 8K and FF would be HD only.
When starting this thread I asked this question to optics experts. The reason was that the answer is really complicated.
Thank our colleague for referring us through a link, I will study the document later.
Best regards

Argyris Theos, gsc
DoP, Athens Greece,
theos@...
+306944725315
Skype Argyris.Theos
www.vimeo.com/argyristheos
via iPhone

31 Μαρ 2018, 3:18 μ.μ., ο/η "Michael Nagel via Cml.News" <michanagel=yahoo.com@...> έγραψε:

CoC calculation, AFAIK, has to do with size of image plane diagonal, not with resolution.

Thomas Gleeson
 

Argyris,

I must admit I don’t fully understand the nature of your question but when calculating the stop difference to match DOF on different size sensors assuming the same aspect ratio you can use a simple formula. In our tests we used the 8k setting on the Monstro at 40.96mm and the 4.5K setting which is 23.04mm. Divide 40.96 by 23.04 and you get 1.78 stops. You could also use the diagonal measurement.

Back to your question I imagine you are doing a thought experiment but in the real world nobody will ever build a HD or  1920x1080 FF35 sensor. This would be a pretty abysmal imager. Even back in the Dark Ages ten years ago did anybody ever build a 1920x1080 single chip camera? I know I am going off on a tangent but even the cheapest S35 cameras have relatively high resolution chips. These machines may have limited processing and recording so you cannot access the full resolution but only the downsampled version at 1920x1080.   

I can recommend this rather long video https://vimeo.com/248235757 from Panavision that discusses many of these topics including the value of capture resolution even over presentation resolution.

Tom Gleeson
Sydney DOP






On 31 Mar 2018, at 8:28 pm, Argyris_Theos_cml <cml@...> wrote:

A question addressing lens experts:
 
Suppose we have two cameras: one carrying S35 8K sensor, the other is FF35 with an HD sensor
I  employ a ~50mm lens on the first, a ~75 on the other. (I choose the exact focal lengths to match field of view)
I place my object at 3 meters and want my Depth of Field to match in both cameras, holding focus from 2.5 to 4 meters. 
The shot wiill be screened in a 10 meter wide screen, via a digital 8K projector. 
The only  post production work allowed is minimal color correction and uprezing of the HD image to 8k at the final stage, in order to avoid using the scaling machine of the projector server. 
Question:
At what f stop do I need to shoot on each camera to match DoF ?
 
Best Regards

Argyris Theos, gsc
DoP Athens Greece
cml@...
skype Argyris Theos

Mike Nagel
 

mine was based on the Zeiss formula, which relates to DOF calculations - which is what you asked about, see here: https://en.wikipedia.org/wiki/Zeiss_formula

again, CoC cannot be based on resolution (as you stated), as film has no resolution - but based on size (as you now seem to state yourself in your last post).

Am I understanding you correctly that you assume the CoC changes based on the final display format ? Or why do you keep mentioning "screen size" ?

Only the capture medium is important here... FF vs. S35

In any case, please keep us updated on your findings, interesting stuff  ;-)

- Mike Nagel
Director/Producer
L.A.

Florian Stadler
 

I used P-Cam and using an Epic Weapon in S35-8K HD (16:9) it prescribes a CoC of 0.00075

To get the Depth of field required you would need an 11 (slightly under)

For the FF HD I used a Canon 5D Mark II cropped to 1.78. P-Cam uses 0.00114 as CoC, needing a 16 stop to get your depth of field.

Florian Stadler
323-377 2242 

On Mar 31, 2018, at 4:04 AM, Geoff Boyle <geoff.cml@...> wrote:

In the lens comparisons I used 2 stop's between FF & S35
Whilst not exactly right it works well with the built-in ND's
Oh and posting which lens you prefer here will have no effect on the outcome which has been gradually changing.
Still need more votes otherwise the only way to find out which is which is to ask Carey at NAB as he's the only other person who knows.

Cheers
Geoff Boyle NSC
Cinematographer
EU based
+31 (0)637155076

Florian Stadler
 

I used P-Cam and using an Epic Weapon in S35-8K HD (16:9) it prescribes a CoC of 0.00075

To get the Depth of field required you would need an 11 (slightly under)

For the FF HD I used a Canon 5D Mark II cropped to 1.78. P-Cam uses 0.00114 as CoC, needing a 16 stop to get your depth of field.

Florian Stadler, DP, LA
www.florianstadler.com

Mark H. Weingartner
 


 
Suppose we have two cameras: one carrying S35 8K sensor, the other is FF35 with an HD sensor
I  employ a ~50mm lens on the first, a ~75 on the other. (I choose the exact focal lengths to match field of view)
I place my object at 3 meters and want my Depth of Field to match in both cameras, holding focus from 2.5 to 4 meters. 
The shot wiill be screened in a 10 meter wide screen, via a digital 8K projector. 
The only  post production work allowed is minimal color correction and uprezing of the HD image to 8k at the final stage, in order to avoid using the scaling machine of the projector server. 
Question:
At what f stop do I need to shoot on each camera to match DoF ?


As an aesthetic exercise this would depend a little on the two cameras but  as a technical exercise, i worked this through comparing 
a RED Weapon with Helium 8k sensor  compared to Monstro 8kVV sensor … all other things are not equal of course, but
as a starting point:

Horizontal angle of view match to a 50 on the S35(ish) sensor is 68.5 on the Weapon 8k VistaVision (which is not really VistaVision)
Using pCAM 

Using pCAM I worked the focus split for the two sensor/lens combinations.  I somewhat arbitrarily set the circle of confusion to the same 0.00049”…  to see what an “apples to apples” comparison would be based on the projector resolution( magnification) as being the limiting factor.

This gave me f/16 for the  S35 and f/32 1/2 for the VV sensor…
Now , going back to an F35 which has a default setting of .00067  pCAM suggests f/16 1/4  …

Obviously all other things are never equal - I would argue that the differences in acuity from different debayer choices and different manufacturers and different lens series makes this a “guestimate” at best… because perceived depth of field in an image is dependent on so many factors… but that is how I would go about the exercise.

We’ve been using circle of confusion as our “fudge factor” much more (i think) in the digital world than when we were looking at film stocks of  a consistent grain size (ish)  projected on a screen of a standard size (sort of)   so that circle of confusion was really varying more with regard to degree of magnification on projection than any other single factor

Weingartner
LA


 
Best Regards

Argyris Theos, gsc
DoP Athens Greece
cml@...
skype Argyris Theos

Pierre Hugues Routhier
 

I think it should be mentioned that closing the iris that much will lead to diffraction -
depending on the lens, you may actually lose what you gained in DoF and end up
with a very soft image.

Pete Routhier, Eng., M.Eng.
Advanced imaging specialist
L.A. and Montreal


On Mar 31, 2018, at 12:11 PM, Mark H. Weingartner <vfxmark@...> wrote:


 
Suppose we have two cameras: one carrying S35 8K sensor, the other is FF35 with an HD sensor
I  employ a ~50mm lens on the first, a ~75 on the other. (I choose the exact focal lengths to match field of view)
I place my object at 3 meters and want my Depth of Field to match in both cameras, holding focus from 2.5 to 4 meters. 
The shot wiill be screened in a 10 meter wide screen, via a digital 8K projector. 
The only  post production work allowed is minimal color correction and uprezing of the HD image to 8k at the final stage, in order to avoid using the scaling machine of the projector server. 
Question:
At what f stop do I need to shoot on each camera to match DoF ?


As an aesthetic exercise this would depend a little on the two cameras but  as a technical exercise, i worked this through comparing 
a RED Weapon with Helium 8k sensor  compared to Monstro 8kVV sensor … all other things are not equal of course, but
as a starting point:

Horizontal angle of view match to a 50 on the S35(ish) sensor is 68.5 on the Weapon 8k VistaVision (which is not really VistaVision)
Using pCAM 

Using pCAM I worked the focus split for the two sensor/lens combinations.  I somewhat arbitrarily set the circle of confusion to the same 0.00049”…  to see what an “apples to apples” comparison would be based on the projector resolution( magnification) as being the limiting factor.

This gave me f/16 for the  S35 and f/32 1/2 for the VV sensor…
Now , going back to an F35 which has a default setting of .00067  pCAM suggests f/16 1/4  …

Obviously all other things are never equal - I would argue that the differences in acuity from different debayer choices and different manufacturers and different lens series makes this a “guestimate” at best… because perceived depth of field in an image is dependent on so many factors… but that is how I would go about the exercise.

We’ve been using circle of confusion as our “fudge factor” much more (i think) in the digital world than when we were looking at film stocks of  a consistent grain size (ish)  projected on a screen of a standard size (sort of)   so that circle of confusion was really varying more with regard to degree of magnification on projection than any other single factor

Weingartner
LA


 
Best Regards

Argyris Theos, gsc
DoP Athens Greece
cml@...
skype Argyris Theos

Mark H. Weingartner
 

True - depending on the lens design of course… but I didn’t set the exercise

I also did not go through the process I should do to adjust the circles of confusion properly - 

This is actually a much more complicated exercise because of the “all other things being equal” issue…

Since our perception of focus is tied to contrast  we really have to consider the mtf of two entire systems…

one from lens to sensor to image processing to projection for the S35 1920x1080 example, and
one from lens to sensor to image processing to projection for the VV 8k example

So many variables in the real world.

Steve Yedlin’s piece probably gives the most concise way of getting from A to B as a practical exercise… it’s pretty close to what I do when shooting, but I have to tweak things frequently when shooting vfx elements because I am rarely shooting an entire frame - more likely an element that does not encompass the entire depth of the scene.
Weingartner
LA

On 31Mar, 2018, at 10:12 49, Pierre Hugues Routhier <pierre.routhier@...> wrote:

I think it should be mentioned that closing the iris that much will lead to diffraction -
depending on the lens, you may actually lose what you gained in DoF and end up
with a very soft image.

Pete Routhier, Eng., M.Eng.
Advanced imaging specialist
L.A. and Montreal


On Mar 31, 2018, at 12:11 PM, Mark H. Weingartner <vfxmark@...> wrote:


 
Suppose we have two cameras: one carrying S35 8K sensor, the other is FF35 with an HD sensor
I  employ a ~50mm lens on the first, a ~75 on the other. (I choose the exact focal lengths to match field of view)
I place my object at 3 meters and want my Depth of Field to match in both cameras, holding focus from 2.5 to 4 meters. 
The shot wiill be screened in a 10 meter wide screen, via a digital 8K projector. 
The only  post production work allowed is minimal color correction and uprezing of the HD image to 8k at the final stage, in order to avoid using the scaling machine of the projector server. 
Question:
At what f stop do I need to shoot on each camera to match DoF ?


As an aesthetic exercise this would depend a little on the two cameras but  as a technical exercise, i worked this through comparing 
a RED Weapon with Helium 8k sensor  compared to Monstro 8kVV sensor … all other things are not equal of course, but
as a starting point:

Horizontal angle of view match to a 50 on the S35(ish) sensor is 68.5 on the Weapon 8k VistaVision (which is not really VistaVision)
Using pCAM 

Using pCAM I worked the focus split for the two sensor/lens combinations.  I somewhat arbitrarily set the circle of confusion to the same 0.00049”…  to see what an “apples to apples” comparison would be based on the projector resolution( magnification) as being the limiting factor.

This gave me f/16 for the  S35 and f/32 1/2 for the VV sensor…
Now , going back to an F35 which has a default setting of .00067  pCAM suggests f/16 1/4  …

Obviously all other things are never equal - I would argue that the differences in acuity from different debayer choices and different manufacturers and different lens series makes this a “guestimate” at best… because perceived depth of field in an image is dependent on so many factors… but that is how I would go about the exercise.

We’ve been using circle of confusion as our “fudge factor” much more (i think) in the digital world than when we were looking at film stocks of  a consistent grain size (ish)  projected on a screen of a standard size (sort of)   so that circle of confusion was really varying more with regard to degree of magnification on projection than any other single factor

Weingartner
LA


 
Best Regards

Argyris Theos, gsc
DoP Athens Greece
cml@...
skype Argyris Theos


Gavin Greenwalt
 

“I want my Depth of Field to match“ \ “The point is that you need different CoC because you have a different native resolution.” - Argyris

This really emphasizes the need to understand a question fully to provide the ‘right’ answer.  In this case the word “match” is ambiguous and needs a little more clarification before you can get the answer you’re looking for.

From an aesthetic perspective what’s most important in matching two plates (say a VV green screen shot and a location s35 shot) is the overall image and you shouldn’t be taking resolution, only the sensor size, into account to match by eye.  If you are compositing multiple plates one would want the foreground  and background defocused areas to look equally out of focus (proportional to the frame) in both shots, otherwise one actor will be incredibly out of focus while the person they’re talking to is perceptually much more in focus.  You want the circl/bokeh to “match” in frame for two things at the same distance. But you also asked for the effective focus region to match.  That’s a different “match” entirely and with the specs you listed would be mutually incompatible with the first form of “matching”.

Without knowing what “matching DOF” means to you in the context of this question you’re going to get two very different answers.  You’ll need to scale the CoC by gate\sensor size alone to match visually.  You’ll need to scale the CoC by resolution and sensor size to match effective focus range (but accept that your images with any subject matter outside of the effective focus region will look very different).  

I can recommend this rather long video https://vimeo.com/248235757 from Panavision” – Tom Gleeson

I would not recommend that video without a list of caveats.   For instance they used sensor plane position instead of Entrance Pupil position when comparing perspective and as a result ended up with differing perspectives which they incorrectly attributed to the format change.   It perpetuates the stubborn myth that faces look inherently different on different sensor sizes (which if true would have meant that decades of Vistavision VFX plates shouldn’t have matched.)  It should be watched with the full knowledge that’s it’s more of a marketing than educational video.

Gavin Greenwalt
VFX Supervisor
Seattle, WA

Leonard Levy
 

First I want to say that I would like to thank the OP for posting a very interesting question that raises a host of issues that don’t seem to have been resolved to me in the many worthy responses I’ve seen so far.
I have recently been wondering just what CoC I should be using to calculate depth of field in video and it seems there are at least 3 variables.

1- The final screen display size - 
I’m guessing this is the most important as I imagine ( though I could be wrong) that classic 35mm DOF tables were based on typical cinematic projection. 
If I am shooting on a Monster but releasing on a web page ( as is more often the case than any of us want to admit) I should have a hell of a lot more perceived  DOF than if its going to end up on an IMAX screen.
So just calculating a DOF without knowing the screen size seems  pointless. 

2- The size of the sensor - 
This would classically have been the critical factor since say in 16mm we needed a much wider lens than in 35 for the same DOF but frankly I can’t recall if there was much difference between DOF in 16mm or 35 with the same lens and F stop,.
In my distant memory through I seem to recall that even here screen size was important and that 16 was assumed to be projected on smaller screens (that’s just a guess though)

3- Pixel density - well that’s a new one isn’t it. Does that affect anything?

I would love to see a clear discussion of how these  parameters ( or any others) weigh against each other.

Leonard Levy, DP
San Rafael, CA




Mitch Gross
 

Screen size used to be the shorthand measure as there was an assumed resolving power & magnification of film projection and another much lower resolving power & magnification for what was then broadcast SD TV. This has changed now, as people can watch a curved 40” 4K computer monitor at home from a distance of less than two feet. This will show higher acuity than a decent movie projection from the front row. 

An from the original question, the idea that a FF-captured HD image will be uprezzed somehow to 8K projection really throws everything out the window. Uprezzed how? Will there be some dithering and smoothing or will pixels simply be repeated to fill out the resolution? The moment this was introduced to the mix it pretty much tossed the comparison out on its ear. 

As to one person’s question if any company had ever actually bothered making a Full Frame 35mm chip that’s only HD resolution, the answer is yes. Vision Research had some super high speed industrial cameras with enormous photosites and that oddball low light Canon is FFHD. 


Mitch Gross
Cinema Product Manager 
Panasonic Media Entertainment Company
New York

On Mar 31, 2018, at 4:22 PM, Leonard Levy <nsll@...> wrote:

First I want to say that I would like to thank the OP for posting a very interesting question that raises a host of issues that don’t seem to have been resolved to me in the many worthy responses I’ve seen so far.
I have recently been wondering just what CoC I should be using to calculate depth of field in video and it seems there are at least 3 variables.

1- The final screen display size - 
I’m guessing this is the most important as I imagine ( though I could be wrong) that classic 35mm DOF tables were based on typical cinematic projection. 
If I am shooting on a Monster but releasing on a web page ( as is more often the case than any of us want to admit) I should have a hell of a lot more perceived  DOF than if its going to end up on an IMAX screen.
So just calculating a DOF without knowing the screen size seems  pointless. 

2- The size of the sensor - 
This would classically have been the critical factor since say in 16mm we needed a much wider lens than in 35 for the same DOF but frankly I can’t recall if there was much difference between DOF in 16mm or 35 with the same lens and F stop,.
In my distant memory through I seem to recall that even here screen size was important and that 16 was assumed to be projected on smaller screens (that’s just a guess though)

3- Pixel density - well that’s a new one isn’t it. Does that affect anything?

I would love to see a clear discussion of how these  parameters ( or any others) weigh against each other.

Leonard Levy, DP
San Rafael, CA




Mike Nagel
 

On Sat, Mar 31, 2018 at 01:22 pm, Leonard Levy wrote:
If I am shooting on a Monster but releasing on a web page ( as is more often the case than any of us want to admit) I should have a hell of a lot more perceived  DOF than if its going to end up on an IMAX screen.
Nobody knows the final delivery "screen size", since there are too many and you cannot control how the audience watches your content. In addition, IF (!) - and that is a BIG IF - there would be a drastic shift/change in perceived DOF based on the viewing medium, then that would not only encompass screen size but also screen technology, as Mitch already partially mentioned. And there are plenty of different screen technologies, and then manufacturer deviations., etc etc etc... And then there is viewing environment, which obviously also has an impact.

I would like to see the equation for all of that.