base-to-base mounting for Sony RX-100 #twinrig #sony3drigs


Oktay
 

I don't want to sound too scientific on an artistic subject but compared to 2D photography 3D photography requires the knowledge of some basic notions.

What I originally tried to comment on was based on the following scenario:

A photographer wants to shoot a 3D photograph of a certain scenery with a horizontally mounted rig and she/he finds out that the minimum stereo base permitted because of the physical dimensions of the cameras is too large considering the position of the nearest and furthest objects.

To counter this problem she/he decides to mount the cameras vertically, because with vertical orientation the stereo base is reduced to the value required by the positions of the objects and the chosen focal length.

I interfere here and say that reducing the stereo base this way does not solve the problem, because the magnification in the equation has also increased because of the decrease in the image width.

I am not advocating here that 3D shooters should limit themselves with strict formulas, but in the above mentioned particular case we are talking about a technical matter that is defined by  stereoscopic rules. Since the desire to seek a smaller base is a technical matter I have commented on the technical side of the subject. 

If a photographer choses to shoot from the ground level, or tries different perspectives or crops part of the image or prefers portrait format with different bases or focal lengths is something that concerns the photographers choices and practices, upon which I find no right in myself to even comment on such choices, styles and practices.

Oktay


George Themelis
 

Wow, I don’t know where to start.

 

Basically, you are saying:

 

  • The only reason to have the cameras vertically should be to achieve portrait orientation, no other reason.
  • If you are concerned about the deviation, as a careful stereo photographer you should adjust the distances of the near/far objects to achieve the desired deviation, not change the stereo base.
  • You should not use cropping to remove near objects, you should not include these objects in the first place.


I don’t know what world you live in, but that’s not my world.

 

Let’s take the RX100 rig as an example:

  • Side-by-side: 112mm stereo base
  • Vertical: 60mm stereo base

 

Are you telling me that pictures taken with 112mm stereo base look the same as pictures taken with 60mm stereo base, as long as the deviation is the same?

 

Absolutely not. There is more in stereo than deviation. The entire perception of the scene & composition is different. If I have a choice, I will go with the normal stereo base for general stereo photography. So, here is a reason to use the vertical orientation: To get “normal” lens spacing.

 

When I go out running with Vinnny, you think I take with me a measuring tape and try to measure every scene before taking a picture? Or, you think I can instruct Vinny to stay here, sit, look straight at the camera or to the side, smile, don’t move, jump over this log now, etc.?

 

No, I take whatever I can with a dog that moves unpredictably around. I go down to Vinny’s eye level and take pictures and hope that something good will come out. There are many things close to me that I cannot control. The 112mm stereo base would have been be a disaster in this situation. Staggered cameras with 75mm stereo base and zoom works better. 60mm or 40mm (Sony RX0) or 30mm (Panasonic 3D1) stereo base works even better.

 

When I use the RX0, I have two choices:

  • Side-by-side: 60mm
  • Vertical: 42mm

 

I do both, but when running with Vinny I go with vertical because I like the reduced stereo base for close-ups. I have no choice with focal length (24mm equivalent). I do a lot of cropping of the foreground which “happens” to be there when I shoot from Vinny’s eye level (or even my eye level) with cameras leveled. How can I avoid the foreground if I want to shoot at Vinny’s eye level looking straight when we are in a narrow trail (grass, etc. everywhere). The only way to avoid the near objects would be to shoot from as high as possible, looking down. But that’s a totally different composition.

 

When I use the RX10 rig (large cameras, 24-600mm zoom), I also mount the cameras vertically. In this case I do it because I want to see through both view-finders to compose in 3D (here is another reason to use vertical cameras).

 

To summarize:

 

  • I use vertical orientation to reduce the stereo base, closer to the spacing of the eyes, or even smaller, for close-ups (or to use the viewfinders with the RX10).

 

  • The portrait orientation is a side-effect of using the cameras vertically and I take care of it by cropping. Sometimes I like the portrait orientation but in most cases I crop to square/horizontal. I lose resolution and field of view (neither is very important for me) but I get what want.

 

  • The “careful stereo photographer” is an ideal concept that does not apply for me and the kind of pictures I like to take. This is photography, not science. Composition comes first. If I can get the composition I want and then crop the near objects to reduce deviation, I will do that.

 

George

 

 

 

 

 

 

From: Oktay via groups.io
Sent: Friday, June 18, 2021 10:14 AM
To: main@photo-3d.groups.io
Subject: Re: [Photo-3d] base-to-base mounting for Sony RX-100 #sony3drigs #twinrig

 

On Thu, Jun 17, 2021 at 08:31 PM, George Themelis wrote:

By the way, cropping and enlarging does not necessarily increase the deviation. This is the case if the distances to the near and far objects do not change. If cropping removes the near object, then you have a different situation. It is possible that the deviation does not change if you do that.

 

George

Presumably the cameras were mounted vertically base to base to decrease the lens separation, because there was a near object that otherwise would have created a problem with the targeted deviation.

A careful stereo photographer who changes the orientation of his cameras due to some stereoscopic concerns would not include into the frame any object  closer than the previously intended,  so that later she/he doesn't need to crop the bottom.

What happens in reality, say with a 3:2 proportion sensor is -compared to the horizontal configuration - you'll have to set the new near object distance 1.5 times further away if you keep the same separation.

If you have reduced the separation by mounting both cameras vertically, your new near object distance will be almost the same (depending how close you could bring the cameras together) as you were using the pair in horizontal configuration, provided that you want to keep the sama deviation value for your viewing preferences.

So, mounting cameras vertically and closer does not solve the problem resulting from including closer objects.
The advantage in the case of mounting cameras vertically would be, when the scene or the objects require a portrait orientation view.

Oktay 

 


Oktay
 

On Thu, Jun 17, 2021 at 08:31 PM, George Themelis wrote:

By the way, cropping and enlarging does not necessarily increase the deviation. This is the case if the distances to the near and far objects do not change. If cropping removes the near object, then you have a different situation. It is possible that the deviation does not change if you do that.

 

George

Presumably the cameras were mounted vertically base to base to decrease the lens separation, because there was a near object that otherwise would have created a problem with the targeted deviation.

A careful stereo photographer who changes the orientation of his cameras due to some stereoscopic concerns would not include into the frame any object  closer than the previously intended,  so that later she/he doesn't need to crop the bottom.

What happens in reality, say with a 3:2 proportion sensor is -compared to the horizontal configuration - you'll have to set the new near object distance 1.5 times further away if you keep the same separation.

If you have reduced the separation by mounting both cameras vertically, your new near object distance will be almost the same (depending how close you could bring the cameras together) as you were using the pair in horizontal configuration, provided that you want to keep the sama deviation value for your viewing preferences.

So, mounting cameras vertically and closer does not solve the problem resulting from including closer objects.
The advantage in the case of mounting cameras vertically would be, when the scene or the objects require a portrait orientation view.

Oktay 


George Themelis
 

Yes, you are approaching this the correct way.

 

Here is an example from my own shooting: When I use my phone to take single camera stereos, I use an app that calculates the deviation. If I want to end up with stereo pairs with about 3% deviation, I aim for something smaller than that (2% or 2.5%) because I know that, after cropping, the deviation will increase.

 

George

 

From: David_in_VA
Sent: Thursday, June 17, 2021 11:03 PM
To: main@photo-3d.groups.io
Subject: Re: [Photo-3d] base-to-base mounting for Sony RX-100 #sony3drigs #twinrig

 

Thank you, George.  That helps a lot.  See if this makes sense:  The "rule of 30" says that if the scene includes an infinity point, the nearest object should be (at least) 30 times the distance between my cameras.  But if I know that I am going to enlarge the stereo image, then I should increase the near-object distance (when I take the photo) to 30 times the new stereo base value as calculated from the formula (including the enlargement factor, or the equivalent increased focal length).  Alternatively, I can crop off the nearer objects until the distance to the nearest remaining object is 30 times the new stereo base.  Is that a valid use of the formula?

 


David_in_VA
 

Thank you, George.  That helps a lot.  See if this makes sense:  The "rule of 30" says that if the scene includes an infinity point, the nearest object should be (at least) 30 times the distance between my cameras.  But if I know that I am going to enlarge the stereo image, then I should increase the near-object distance (when I take the photo) to 30 times the new stereo base value as calculated from the formula (including the enlargement factor, or the equivalent increased focal length).  Alternatively, I can crop off the nearer objects until the distance to the nearest remaining object is 30 times the new stereo base.  Is that a valid use of the formula?


George Themelis
 

“If I take a stereo picture with two cameras, the distance between the lenses is called the "stereo base", isn't it? “

 

Yes.

 

“If I then crop/enlarge, and put the new larger deviations into the formula, I will get a different value for "stereo base" (probably larger). “

 

How are you using this formula?  If you are using it to calculate the stereo base that will give you a certain deviation, then after you crop and enlarge the picture, if you want to end up with the same deviation (which is usually the goal), you need to use a smaller stereo base. That’s what the formula tells you.

 

It boils down to the question: What is your goal?

 

“If I worked very hard to put the cameras very close together (to approximate the human interocular distance), does the new larger value mean that I wasted my time?”

 

Was your goal to approximate the human interocular distance or to achieve a certain deviation? If your goal was to approximate the human interocular distance, then you are done. You control the deviation by adjusting the distance to the nearest object (usually).

 

By the way, cropping and enlarging does not necessarily increase the deviation. This is the case if the distances to the near and far objects do not change. If cropping removes the near object, then you have a different situation. It is possible that the deviation does not change if you do that.

 

George

 

 

 

From: David_in_VA
Sent: Thursday, June 17, 2021 8:22 PM
To: main@photo-3d.groups.io
Subject: Re: [Photo-3d] base-to-base mounting for Sony RX-100 #sony3drigs #twinrig

 

Thanks, John.  I understand that the deviations will increase proportionally when you zoom or crop.  That's an interesting technique for making a miniature look full sized.  I'll have to try it.

But I apologize for not being more clear in my question.  My confusion is that the term "stereo base" seems to be used in more than one way.  If I take a stereo picture with two cameras, the distance between the lenses is called the "stereo base", isn't it?  If I then crop/enlarge, and put the new larger deviations into the formula, I will get a different value for "stereo base" (probably larger).  What is this new value telling me, that the old one did not?  If I worked very hard to put the cameras very close together (to approximate the human interocular distance), does the new larger value mean that I wasted my time?  (I don't think so, but again, I must be missing something.)

 


David_in_VA
 

Thanks, John.  I understand that the deviations will increase proportionally when you zoom or crop.  That's an interesting technique for making a miniature look full sized.  I'll have to try it.

But I apologize for not being more clear in my question.  My confusion is that the term "stereo base" seems to be used in more than one way.  If I take a stereo picture with two cameras, the distance between the lenses is called the "stereo base", isn't it?  If I then crop/enlarge, and put the new larger deviations into the formula, I will get a different value for "stereo base" (probably larger).  What is this new value telling me, that the old one did not?  If I worked very hard to put the cameras very close together (to approximate the human interocular distance), does the new larger value mean that I wasted my time?  (I don't think so, but again, I must be missing something.)


John Clement
 

Cropping does make changes in the image, which is also significant.  A cropped image has an effect the same as a zoom.  The far points are more widely separated on your display device, assuming there is no adjustment of the R/L alignment.  The cheap way to make a miniature look full size is to take it from a distance and then crop, or zoom in.  It of course is not quite right, but it is much easier than using a slide bar to have effective lens separation smaller than normal with the camera close to the object.  No matter how you do it, the results will look different depending on the display devices sizes.

 

John M. Clement

 

From: main@Photo-3d.groups.io <main@Photo-3d.groups.io> On Behalf Of David_in_VA
Sent: Thursday, June 17, 2021 10:30 AM
To: main@Photo-3d.groups.io
Subject: Re: [Photo-3d] base-to-base mounting for Sony RX-100 #sony3drigs #twinrig

 

I understand the math that says the stereo base can be calculated as a function of focal length, deviation, object distance, etc. (Ferwerda page 238; Waack page 68), but how does that actually relate to the stereo image?  Can someone help me understand this?  I must be missing something.  Cropping clearly doesn't change the original camera locations, which I thought were the endpoints of the actual stereo base.  So are we calculating an "effective stereo base"?  If so, what does it mean and how does it affect the stereo image?  Thanks.


David_in_VA
 

I understand the math that says the stereo base can be calculated as a function of focal length, deviation, object distance, etc. (Ferwerda page 238; Waack page 68), but how does that actually relate to the stereo image?  Can someone help me understand this?  I must be missing something.  Cropping clearly doesn't change the original camera locations, which I thought were the endpoints of the actual stereo base.  So are we calculating an "effective stereo base"?  If so, what does it mean and how does it affect the stereo image?  Thanks.


George Themelis
 

Hi Oktay, et al.,

 

That’s a good point, but it needs a clarification.

 

Consider the original image (2:3 aspect ratio) out of the RX100 oriented vertically.

 

The first step is crop. When you crop, two things happen:

  1. You lose resolution (you throw away pixels)
  2. You lose field of view

 

In the attached picture I have cropped the original image into a 4:3 landscape mode. Essentially I threw away half of the image (the bottom half). The way I cropped, I have lost field of view only vertically, not horizontally. The deviation has not changed by cropping, only the resolution and the field of view.

 

The second step is to enlarge the image to fill the display. This is a separate step but when you use SPM to crop, the image is scaled automatically so for most people this happens when you crop, but not necessarily.

 

It is in the step that the stereoscopic deviation increases. Why? Because I have enlarged the image by 2x. I prefer to think of the deviation as a function of magnification and stereo base, not so much the focal length (Deviation = Magnification x Stereo Base)

 

The stereoscopic deviation could have increased by as much as 2x, but it is possible that it is not increased that much. When I cropped the image at the bottom, most likely I have cropped the near-by objects (most likely the ground). The stereoscopic deviation depends on the distances of the near and far objects. If these have remained in the picture when I cropped, then, yes, the deviation has doubled. But if they were cropped out, then the deviation has increased less.

 

I use vertical cameras a lot. For example, my twin Sony RX10 rig. I crop and I use cropping as a way to increase the magnification. I am also using the RX0 rig vertically.  With the RX0, I crop a lot of the ground. I do not mind that the effective focal length increases, I actually like it.

 

I have not tried the RX100 vertically yet, but if you ask me what I prefer, 1) the cameras side-by-side with 112mm separation at 24mm focal length, or 2) the cameras vertically with 60mm separation and focal length of 35 to 50mm, I would say in most cases I would prefer #2.

 

George

 

 

 

 

 

From: Oktay via groups.io
Sent: Saturday, June 12, 2021 6:45 PM
To: main@photo-3d.groups.io
Subject: Re: [Photo-3d] base-to-base mounting for Sony RX-100 #sony3drigs #twinrig

 

If you use cameras in vertical orientation and crop a horizontal rectangle, keep in mind that the effective 35mm equivalent focal length will increase.
So what you gain in base will be lost due to this increase in FL if you want to maintain the same deviation.
In other words you'd need a smaller base than the one you've obtained by mounting the cameras bottom to bottom.

Oktay

 


Oktay
 

If you use cameras in vertical orientation and crop a horizontal rectangle, keep in mind that the effective 35mm equivalent focal length will increase.
So what you gain in base will be lost due to this increase in FL if you want to maintain the same deviation.
In other words you'd need a smaller base than the one you've obtained by mounting the cameras bottom to bottom.

Oktay


Gordon Au
 

FYI, this is my solution for for mounting vertically, but top-to-top, using 2 x RX-100 mark ii. Very few pieces—L-bars and a couple screws easily found online, and can accommodate a handle. I need to measure the interaxial exactly, but I think it's close to what you would get with an Arca-Swiss rail or comparable in between the bottoms. Because the bottom holes are at fixed positions on these L-bars, this is the closest I can get them, but you could close the remaining ~4mm gap using ones with continuous slots.


robert mcafee
 

Get a vertical Arca Swiss rail (double dovetail). Screw one camera to the rail directly. Get an Arca Swiss clamp. Screw camera to clamp. Clamp onto Arca Swiss vertical rail (onto dovetail on opposite side of mounted camera) and adjust position so lens axes coincide. 

Distance between camera bases will be about 5/8” or a bit less. 




On Friday, June 11, 2021, 2:02 PM, timo@... <timo@...> wrote:

Is there a battery door or something, on the bottom of the cameras?

Timo

On Jun 11, 2021, at 1:57 PM, Paul Gillis <pgillis@...> wrote:

The lengthy discussion about getting the lenses of Sony RX-100's close together inspired me to draw up a crude diagram of my idea for mounting such cameras base-to base.  As George noted, the tripod mounting hole is not aligned with the lens, so a single short threaded rod will not align the cameras.  My idea is to mount the cameras base-to-base on a single aluminum bar, using flathead machine screws, their heads countersunk into the bar.  You would position the cameras at a right angle to the bar, insert the screws, and screw them in until almost tight.  Then you would rotate the cameras clockwise into their proper position on the bar, and engage a small clip or retainer at each end of the bar to prevent further camera rotation.

The anti-rotation clips at the ends of the bar would have to be sized/shaped to cradle the end of the camera farthest from the tripod socket.  They could be attached by a screw or just slide onto the bar.  (An even slicker arrangement would be some kind of spring clip.)

Unfortunately, I don't have any cameras with offset tripod sockets to try this out on. (Although I recently acquired some Canon S-95s, whose sockets seem to be offset about 3mm from the lens centerline, too close for this scheme to work.)

Looking at a dimensional illustration of the original RX-100, it seems this arrangement would give a stereo base of about 65 mm, if you used a 1/4" thick mounting bar.  Of course, you need to be willing to take photos in portrait orientation.  If you crop down to a 4:5 horizontal rectangle, you still have over 10 megapixels to work with.

--Paul Gillis

<Offset base-to-base camera bar.jpg>


timo@guildwood.net
 

Is there a battery door or something, on the bottom of the cameras?

Timo

On Jun 11, 2021, at 1:57 PM, Paul Gillis <pgillis@...> wrote:

The lengthy discussion about getting the lenses of Sony RX-100's close together inspired me to draw up a crude diagram of my idea for mounting such cameras base-to base.  As George noted, the tripod mounting hole is not aligned with the lens, so a single short threaded rod will not align the cameras.  My idea is to mount the cameras base-to-base on a single aluminum bar, using flathead machine screws, their heads countersunk into the bar.  You would position the cameras at a right angle to the bar, insert the screws, and screw them in until almost tight.  Then you would rotate the cameras clockwise into their proper position on the bar, and engage a small clip or retainer at each end of the bar to prevent further camera rotation.

The anti-rotation clips at the ends of the bar would have to be sized/shaped to cradle the end of the camera farthest from the tripod socket.  They could be attached by a screw or just slide onto the bar.  (An even slicker arrangement would be some kind of spring clip.)

Unfortunately, I don't have any cameras with offset tripod sockets to try this out on. (Although I recently acquired some Canon S-95s, whose sockets seem to be offset about 3mm from the lens centerline, too close for this scheme to work.)

Looking at a dimensional illustration of the original RX-100, it seems this arrangement would give a stereo base of about 65 mm, if you used a 1/4" thick mounting bar.  Of course, you need to be willing to take photos in portrait orientation.  If you crop down to a 4:5 horizontal rectangle, you still have over 10 megapixels to work with.

--Paul Gillis

<Offset base-to-base camera bar.jpg>


Paul Gillis
 

The lengthy discussion about getting the lenses of Sony RX-100's close together inspired me to draw up a crude diagram of my idea for mounting such cameras base-to base.  As George noted, the tripod mounting hole is not aligned with the lens, so a single short threaded rod will not align the cameras.  My idea is to mount the cameras base-to-base on a single aluminum bar, using flathead machine screws, their heads countersunk into the bar.  You would position the cameras at a right angle to the bar, insert the screws, and screw them in until almost tight.  Then you would rotate the cameras clockwise into their proper position on the bar, and engage a small clip or retainer at each end of the bar to prevent further camera rotation.

The anti-rotation clips at the ends of the bar would have to be sized/shaped to cradle the end of the camera farthest from the tripod socket.  They could be attached by a screw or just slide onto the bar.  (An even slicker arrangement would be some kind of spring clip.)

Unfortunately, I don't have any cameras with offset tripod sockets to try this out on. (Although I recently acquired some Canon S-95s, whose sockets seem to be offset about 3mm from the lens centerline, too close for this scheme to work.)

Looking at a dimensional illustration of the original RX-100, it seems this arrangement would give a stereo base of about 65 mm, if you used a 1/4" thick mounting bar.  Of course, you need to be willing to take photos in portrait orientation.  If you crop down to a 4:5 horizontal rectangle, you still have over 10 megapixels to work with.

--Paul Gillis