Home of Professional Cinematography since 1996

>CCD Flange Focal Distances

Published : 26th August 2003

Just checking in...

Someone explained that CCDs in a color prism block camera are mounted at different distances, optically, from the flange, to permit simpler lens design.

While I know that this was the case with tube cameras, I have trouble believing it's true of CCD cameras. But I'm willing and eager to learn. Does anyone have the standard but different distances (there'd have to be one) that the R, G, and B CCDs are mounted to?

Jeff Kreines



Jeff Kreines wrote:

>Someone explained that CCDs in a color prism block camera are >mounted

I sort of don't think so. Here's an interesting question, are all video lenses (for 3xCCD cameras) achromats ? Are any apochromats? Which are DigiPrimes ?

I think if you use chip position (with respect to focal plane) to compensate for spherical aberration or longitudinal chromatic aberration you then must overcorrect for lateral chromatic aberration (color difference of magnification) because in displacing any chip you are effectively causing a change in image size, right ?

Sam Wells



Jeff Kreines writes:

>Someone explained that CCDs in a color prism block camera are >mounted

I believe it has more to do with the finite math than determines the crossover point of the image than an arbitrary factor such as the distance of a terminal element from the focal plane or a machined element such as a flange. (sorry for the long run-on)

Scott Billups – LA



Jeff Kreines wrote:

>Someone explained that CCDs in a color prism block camera are >mounted

Why would the conversion from tubes to CCD change anything to this concept? When cameras changed from tubes to CCD we continued to use the
same lenses. Optic rules did not change. It continued to be cheaper to produce lenses that way. There was no change made to the established standards for manufacturing television lenses, AFAIK.

Recently there HAS BEEN an opportunity to change standards, and that was HDTV. And they did. But again it was agreed upon to have different distances for R, G and B. Although not the same values as the ones established for SDTV...... (Fweew, I managed to stay on-topic)

>Jeff "needs some evidence" Kreines

I am not in possession of the standards or hard evidence at this time. It was explained to me by a very knowledgeable representative of a major TV lens manufacturer.

Michael Bravin,

>I suspect they studied the case closely.

He confirms this still being the case in HDTV. I don't have any more proof handy at this time, but I suspect this to be true, also because it looks commercially quite viable to me.

Economics often dict technical decisions, certainly in the TV/Video world.

Hope this helps,, best,

Kommer Kleijn
VFXCinematographer

http://www.kommer.com



Kommer Kleijn wrote:

> Yep, I think that someone was me


It was actually someone else, though you may have concurred.

>While I know that this was the case with tube cameras, I have trouble >believing it's true of CCD cameras.

Actually, I recall different lenses being required when CCD cameras came in, which didn't please those upgrading from their Ike HL79Es. With tubes, obviously, everything was in flux, as they couldn't be cemented permanently in position. But CCDs changed that.

I'm still looking for some proof here!

Jeff Kreines



>Actually, I recall different lenses being required when CCD cameras >came in

I also remember this... the lens brochures of the day differentiated between "CCD" and "Tube" lenses. Don't know how much of that was actual science and how much was marketing hype. But the TV station I was working at in 1986 wasn't in the habit of spending money if they didn't have to... and we got new lenses with all of our CCD cameras that year.

George Hupka
Director/DP, Downstream Pictures
Saskatoon, Canada



Mitch Gross wrote :


> A properly designed lens (like the DigiPrime) will focus Red Green and >Blue

If the DigiPrime is as described above it is an apochromat, and should be superior on film, assuming there are no other differences in the design. But I assume that it is also designed to counteract chromatic aberration introduced by the prism block, so would not perform as well on a film camera with no block, at least at wide apertures.

Wade K. Ramsey, DP
Dept. of Cinema & Video Production
Bob Jones University
Greenville, SC 29614


>Pixel Monger wrote:

>I believe it has more to do with the finite math than determines the >crossover point of the image than an arbitrary factor

Yes, I'd agree completely. But the earlier poster implied that the optical (if not physical) flange focal distance for each color was different -- and the implication was that it wasn't merely that different wave lengths of light are of, well, different lengths. The implication was that there was a standard offset of some sort designed into the prism block. Obviously, if such an offset existed, it would have to be standardized between all manufacturers, or lens interchangability would suffer.

My guess is that the CCDs are all either:

a) at the same mechanical distance from the flange

b) at an optically-compensated distance from the flange, so that all three colors now fall into exactly the same plane.

My money is on A, of course. But the Bolex issue arises here, too. The thick prism used in Bolex Rex cameras requires compensation in the optics (see Dennis Couzin's great website for more), and, obviously, a big video camera prism is going to require even more significant compensation (not just back focus).

But there are cameras that used huge prism blocks (CP's SPR reflex conversion of the Mitchell BNC) that apparently used standard BNCR mount lenses that also could be used on a spinning mirror BNCR, i.e. the 25-250. Though, perhaps, the long focal length and slow speed of that lens prevented the prism from being a significant problem.

Perhaps, then, it's the very short focal lengths of lenses used on 2/3" prism cameras that make all of these problems more significant.

Where's David Samuelson, Les Zellan, or Joe Dunton when you need them?

Jeff Kreines



Jeff Kreines wrote:

>But the earlier poster implied that the optical (if not physical) flange focal >distance for each color was different.

Absolutely.

>My guess is that the CCDs are all either…
>a) at the same mechanical distance from the flange


They can't be, due to the different light path, needed for the color separation they are at very different distances, even different angles. Only the green CCD is at the "normal" position. R an B are pretty far out. I'll try to find a schematics of a TV 3CCD prism and post it.

>b) at an optically-compensated distance from the flange, so that all >three colors now fall into exactly the same plane.

This is more like it, except that they fall ALMOST in the same plane optically. There is a very small (standardized) difference, created on purpose to allow for easier (and thus cheaper) lens design.

Best!,

Kommer Kleijn
VFX Cinematographer



Krommer wrote:

>There is a very small (standardized) difference, created on purpose to >allow for easier (and thus cheaper) lens design."

Don't think so It is always MORE expensive to make a lens that properly
compensates for Lateral Chromatic aberration.

Lens design compensating for lateral chromatic aberration is needed to "cancel out" the LCA caused by the prism offsets from the Green to Red To Blue CCD's fixed to the prism.

I will find the correct offsets for HD 2/3" CCD Prism block and post them.

I think it is G=0 Red= +5 microns Blue= +10 microns

A properly designed lens (like the DigiPrime) will focus Red Green and Blue light on EXACTLY the same plane eliminating any discernable Lateral Chromatic aberrations. When you see color fringing especially in the edges of longer lenses it is most probably improperly or un corrected LCA.

Flange distance is to the "0" located or green CCD and it is at 65.06mm (or 48mm in air) corrected for all the glass- Low pass IR Cut and Prism block.

It's difficult and complicated and expensive to design lenses that FULLY OPTIMIZE the image captured at the CCD after the Beam splitting prism and that is why DigiPrimes are considerably more expensive than their film brethren.

It should also be noted that NO CURRENTLY COMMERCIALLY AVAILABLE OR VIABLE PRODUCT for digital cinematographic image capture exceeds the imaging capabilities and performance of the 2/3" CCD Prism Block HD imager with a DigiPrime mounted on it.

Michael Bravin
Chief Technology Officer
Band Pro Film Video, Inc
www.digiprimes.com



Mitch Gross wrote:

>So I guess this brings the question back around to the previous thread; >using HD lenses on a film camera. How would a "properly designed" HD >lens such

In lens design, any planar glass (slab) between the lens and focal plane must be accounted for in the design. Planar glass distorts the focal point such that the outside of the lens focuses deeper than the centre of the lens.

Here's what happens:

Draw a diagram to follow this - (Do attachments go through this list? - it would be easier to show)

Draw a simple lens with a 50mm focal length. Make the lens f1.0 to illustrate the point. (50mm diameter) Draw 3 rays - from the lens to the focal point - one from the edge, one from the centre (the lens axis ) and one half way in between. Now draw a 10mm thick slab of parallel sided glass perpendicular to the lens axis, between the lens and the focal point.

Where outside ray goes through the glass, show some diffraction where the ray bends towards the perpendicular axis on entry to the glass, and back parallel to the incident ray on exit. Continue this exit ray to the lens axis. It crosses the lens axis behind the 50mm focal point. - and represents the focal point for this ray. Where the middle ray (half way between the outside and axis) enters the glass slab, refract it in a similar manner. Note that the angular differences are less. Draw the exit ray from the glass slab parallel to the incident ray. This ray will cross the lens axis closer to the lens than the outside ray.

(Anybody still with me?)

This shows that the focal point will become distorted with a glass slab in between. How significant is this distortion? For a 50mm lens with 10mm of glass (quartz -index = 1.67), the spot focus will be located between 54.46 (for the f1 annular ring) and 54.04 mm (for the centre rays) from the lens (with 50mm being the nominal without glass slab). This results in a blur circle diameter of 0.16mm This difference is very significant for 16 or 35mm film At f2, the difference is less - spot elongation of 0.11mm and blur circle diameter of 0.02mm. It continues to get better with smaller aperture. If you put a 50mm block of glass behind it, the blurring is much worse.

One more level of complexity is added because the colors will refract differently in the glass slab, resulting in R,B,and B focusing at a different point, and resulting lateral color shift which will show up symmetrically at the edges of the image from the lens.

In summary, these issues are accounted for in any lens system which is intended to operate with a slab of glass behind it (whether a prism or a significant thickness of filters.)

Matt Cowan



Kommer Kleijn wrote:

>They can't be, due to the different light path, needed for the color >separation they are at very different distances

I respectfully disagree.

Obviously, the CCDs are at different angles, but all of them end up being optically parallel to the lens mount, or the image would be distorted. As for distance, if the distance for each CCD to the flange varied more than by the difference in the wavelength of the color, magnification would vary. Either of these are serious no-no's.

Though it's a bit hard on the eyes to read the grey over white text...

Jeff Kreines



> I will find the correct offsets for HD 2/3" CCD Prism block and post >them. I think it is G=0 Red= +5 microns Blue= +10 microns

Dejavu...I just saw this in Fujinon's last HD zoom brochure - actually there its +5 microns blue, and +10 microns red.

Mark Doering-Powell
Director of Photography



Michael wrote:

>I will find the correct offsets for HD 2/3" CCD Prism block and post them. >I think it is G=0 Red= +5 microns Blue= +10 microns

Mark wrote :


> Dejavu... I just saw this in Fujinon's last HD zoom brochure - actually >there its +5 microns blue, and +10 microns red.

Makes perfect sense. I was just asking for actual figures.

Though I'd have guessed if it was wavelength based it would be minus
blue, green zero, and plus red...I wonder how closely that all corresponds to the actual distances between film layers?

Where's John Pytlak when you need him?

Jeff Kreines



Jeff Kreines writes:

>Obviously, the CCDs are at different angles, but all of them end up >being optically parallel to the lens mount, or the image would be >distorted

That much is correct.

>As for distance, if the distance for each CCD to the flange varied more >than by the difference in the wavelength of the color, magnification >would vary.

Right. But since red and blue converge (and form an image of the prescribed size) at their own respective distances, there's no magnification discrepancy when the chips are placed that those respective distances. It's as if relay lenses had been placed in the red and blue optical paths. The net result is an image of the prescribed size, but projected at a different distance.

Dan Drasin
Producer/DP
Marin County, CA



Dan Drasin wrote:

>Right. But since red and blue converge (and form an image of the >prescribed size) at their own respective distances,

If you'd quoted more of my message, you'd see I made the same point,
Dan.

Someone had suggested that the distances of the R and B CCDs were radically different than the G -- i.e. not just a few microns. That's what I was contesting.

Jeff Krines



Michael Bravin wrote:

> Lens design compensating for lateral chromatic aberration is needed >to I think it is G=0 Red= +5 microns Blue= +10 microns

Seems since the lens has to focus each primary color to its respective target rather than all three colors converging to a flat pinpoint plane chromatic aberration should be zero. So what's going on with these zooms and blue fringing?

Tom McDonnell
DP
New Orleans, La



> Seems since the lens has to focus each primary color to its respective >target

Well just make sure your subject matter consists only of spectrally pure RGB primaries seriously, an achromatic lens will land color focus correctly at two
wavelengths on the spectrum.

Acceptable aberration would be how much deviation from those points you can live with.

Sam Wells



Tom wrote:

>the lens has to focus each primary color to its respective target...

Tom you make this sound simple inexpensive and common. It is none of these and that is the difference between a great lens and an inexpensive lens and also one of the reasons why film lenses are ill suited for HD as it currently is captured.

Michael Bravin
Chief Technology Officer
Band Pro Film Video, Inc



Michael Bravin wrote:

> Tom you make this sound simple inexpensive and common It is none >of these

Part of the reason (it was a horrible film and I saw it only for technical reasons) that THE REAL CANCUN looked so bad was, I'm sure, the mediocre lenses used to shoot it. Actually painful to watch after a while...

Jeff Krines



Jeff wrote:

>Part of the reason...that THE REAL CANCUN looked so bad was,

May be the lenses, but things like skill of the crew and DP, Proper camera set-up and care taken in Color Correction, Exposure and Film-out were probably bigger factors. If everything else had been done properly the lenses wouldn't have been such a factor.

Jackpot DP David Mullen used an average HD lens and it looked very good.
Michael "believes in great optics but also wants all eight cylinders firing"

Michael Bravin



Michael Bravin wrote:

>Tom you make this sound simple inexpensive…

Michael you misunderstood my statement. I in no way meant to say this was easy or inexpensive. What I meant was since each color has its own target and depth it seems it might be easier to design out chromatic aberration than correct for it later. Easy!, the only thing easy I can think of is getting into debt...

Tom McDonnell
DP
New Orleans, La



A properly designed lens (like the DigiPrime) will focus Red Green and Blue light on EXACTLY the same plane eliminating any discernable lateral chromatic aberrations. When you see color fringing especially in the edges of longer lenses it is most probably improperly or un corrected LCA.

So I guess this brings the question back around to the previous thread; using HD lenses on a film camera. How would a "properly designed" HD lens such as a DigiPrime perform on a Super-16 film camera? The image area isn't much larger, so I'm guessing that many of these lenses would cover. If I understand this thread correctly (and there's a serious chance that I don't) a really good HD lens should deliver a really good image on a film camera.

Anyone know?

Mitch Gross
NYC DP



Mitch wrote:

>If I understand this thread correctly (and there's a serious chance that I >don't) a really good HD lens should deliver a really good image on

Mitch……It looks like you may have missed the point.

HD lenses designed for CCDs and their prisms are corrected for that prism and film cameras have no RGB beamsplitting prism. This is part of the optical design when designing for HD with a prism.

Without the optical correction for the prism you do not have a lens optimised for an HD prism. You also do not have the same lens. Aside from this correction, the MTF is nearly 2 times that found in the best film lenses.

Michael Bravin
Chief Technology Officer
Band Pro Film/Video, Inc



Michael Bravin writes:

>Without the optical correction for the prism you do not have a lens >optimised for an HD prism.

Which begs the question: Is the prism correction accomplished as a matter of overall design, or is there, say, a single rear element that could be swapped with one that's finetuned for film? Seems to me that Zeiss would have done the latter, if possible, to protect against the obsolescence of their product when single-chip HD designs arrive in a few years.

Dan Drasin
Producer/DP
Marin County, CA



Dan wrote :


>…….Seems to me that Zeiss would have done the latter, if possible, to >protect against the obsolescence of their product when single-chip HD >designs arrive in a few years.

Dan

There is no free lunch. You could just add a rear group to handle the prism offsets, some film lens manufacturers do this to protect their film lens business. These lenses cannot and will not perform optimally. The CLA35HD from Angenieux/Zeiss was such a product and it was and is the finest quality relay for HD. But a relay however good is no substitute for a purpose built lens.

Designing a lens "system" that fully optimises the image and the optical properties for HD with a prism requires an overall design with few compromises-Zeiss DigiPrimes. If one as you call it "protect against obsolescence" you may win the an economic battle but you have less of a lens. Zeiss designed the DigiPrimes for a 3 CCD 2/3" HD prism. Putting them on any other system may work but at reduced performance. This is partially why the very best 35mm film lenses are inadequate for HD. Uncompromising image quality for 2/3" 3CCD HD cameras is the game we are leading, please come play.

Michael Bravin
Chief Technology Officer
Band Pro Film/Video, Inc



The Foveon chip set ( www.foveon.com ) being brought out now, promises us the 35mm sized sensor, and pixels with sensitivity to R G and B that are stacked front to back, rather than using prisms or placing RGB pixels next to each other. Does this mean that, should this design live up to its promises, that another lens design will likely be needed to work well with this new chipset?

I saw a comparison from the still photography world.

The Foveon was put up against the high-end Nikon digital, and the Foveon was incredibly superior.

K Stebleton



K Stebleton wrote :


>"The Foveon chip set ( www.foveon.com ) being brought out now, >promises us the 35mm sized sensor..

1. Foveon's web site describes this chip as having a 25mm diagonal, which may not cut it in the future.

2. Where did you see the Nikon/Foveon comparison?

Jerry Cotts
DP/LA



K Stebleton wrote:

>The Foveon chip set (www.foveon.com ) being brought out now, promises us the 35mm sized sensor, and pixels with sensitivity to R G and B

But it can only run at about 2 frames per second.

Optically, it will be the same as a single CCD or film camera, as the distance between color planes is pretty negligible. No prism.

Jeff Kreines