When small interaxials just don't cut it! - Page 22 - AVS Forum
Forum Jump: 
Reply
 
Thread Tools
post #631 of 1097 Old 05-16-2012, 03:43 AM
AVS Club Gold
 
Don Landis's Avatar
 
Join Date: Jun 1999
Location: Jacksonville, FL
Posts: 10,761
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 37 Post(s)
Liked: 106
Quote:


But if this is true it should be possible to increase the heigth of one camcorder a little bit, by laying something between the igus and the ballheads.

Yes, that's what my shims do. Use the grid cross hairs on your TD10 screens to adjust the vertical disparity. I think those ball heads will become a "monkey on your back" . Nobody uses them. I even got rid of my set screws in my final design.


Further thought on your PM question. When I do my alignment, I do the coarse adjustment of the parallel using wide angle. Then I do the fine adjust with the full Zoom at 30 meters on a target. But my project shooting is mostly done between full wide and only about half the full zoom or about 5:1. Doing this is not much different than using a laser bore sight to align a scope on a sniper rifle. But in this case you are not adjusting an angle for convergence on the target in the reticule but maintaining a parallel with the two TD10's.
Don Landis is offline  
Sponsored Links
Advertisement
 
post #632 of 1097 Old 05-16-2012, 04:43 AM - Thread Starter
AVS Special Member
 
Frank's Avatar
 
Join Date: Apr 1999
Location: Dept. of Offense
Posts: 5,194
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 4 Post(s)
Liked: 14
Wolfgang, I have been assuming you have the OIS disabled for both TD-10s
when using them in a dual camera setup.
You do, right?

Over the last several years I experimented with over a dozen different
combination of camcorders in dual cam setups.
I have never found two cameras that had perfect alignment throughout the
full zoom range. The two Canon HF-M32s that I used in the YouTube zoom
test are as close as any I've ever tested.
Considering the complexity of the optical system which includes a
OIS I can't expect better.

Did IQ's suddenly drop sharply while I was away?
I enjoy 3D in spite of HDMI 1.4!
Full screen only 3D doesn't cut it!
Frank is offline  
post #633 of 1097 Old 05-16-2012, 05:20 AM
AVS Special Member
 
Wolfgang S.'s Avatar
 
Join Date: Aug 2011
Location: Vienna/Austria
Posts: 1,131
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 5 Post(s)
Liked: 16
Quote:
Originally Posted by Don Landis View Post

Yes, that's what my shims do. Use the grid cross hairs on your TD10 screens to adjust the vertical disparity.

Well, I use the cross hairs to adjust both the horizontal and vertcal disparity.


Quote:
Originally Posted by Don Landis View Post

I think those ball heads will become a "monkey on your back" . Nobody uses them. I even got rid of my set screws in my final design.

Maybe you are right. But at least I see here a good adjustment possibility. But I have also received this morning the Manfrotto 323 that you use - and will be keen to see how they come out and then I am able to compare the different methods.


Quote:
Originally Posted by Don Landis View Post

Further thought on your PM question. When I do my alignment, I do the coarse adjustment of the parallel using wide angle. Then I do the fine adjust with the full Zoom at 30 meters on a target. But my project shooting is mostly done between full wide and only about half the full zoom or about 5:1. Doing this is not much different than using a laser bore sight to align a scope on a sniper rifle. But in this case you are not adjusting an angle for convergence on the target in the reticule but maintaining a parallel with the two TD10's.

Yes, up to now I am going for parallel axis. But there are limitations that I have become aware with the first tests - if the IO becomes to large, Vegas is not able any more to adjust the parallel axis to move everything behind the stereo window. So I think that I will tend to go more for far point parallax, especially with larger IOs.

Well, the first allignemt was done with full zoom at 20 m distance - and yes, it was done in a parallel way. But then I have done my first test shootings in full wide angel, and detected here that the vertical disparity is much too high.

In the next step I did the adjustment in full wide angel - and had better results since I shoot in full wide angel. So my next test is to see if needs an better adjustment in the heigth.


Quote:
Originally Posted by Frank View Post

Wolfgang, I have been assuming you have the OIS disabled for both TD-10s when using them in a dual camera setup.
You do, right?

Frankly spoken, I do not. But you are right, that is something that I have to do. Most of my testing up to now were statical pictures, but it makes no sense to run the OIS if you use a tripod. I simply forgot that since I do not use tripods a lot of time up to now. I will disable that OIS and will be interesting to see if that changes things dramatically.


Quote:
Originally Posted by Frank View Post

I have never found two cameras that had perfect alignment throughout the full zoom range. The two Canon HF-M32s that I used in the YouTube zoom test are as close as any I've ever tested.

Well, I am willing to accept some deviation, but what I see here is too much. At least that is my interpretation.

Kind regards,
Wolfgang
videotreffpunkt.com
Wolfgang S. is offline  
post #634 of 1097 Old 05-16-2012, 06:01 AM - Thread Starter
AVS Special Member
 
Frank's Avatar
 
Join Date: Apr 1999
Location: Dept. of Offense
Posts: 5,194
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 4 Post(s)
Liked: 14
Quote:
Originally Posted by Wolfgang S. View Post

Frankly spoken, I do not. But you are right, that is something that I have to do. Most of my testing up to now were statical pictures, but it makes no sense to run the OIS if you use a tripod. I simply forgot that since I do not use tripods a lot of time up to now. I will disable that OIS and will be interesting to see if that changes things dramatically.




Well, I am willing to accept some deviation, but what I see here is too much. At least that is my interpretation.

I suspected you might have the OIS enabled.
IF you don't see a dramatic improvement after disabling it, I would be extremely surprised.

I learned the hard way several years ago to never ever enable OIS when shooting twin camcorder 3D!
Even if the cameras are fixed on a tripod the camcorder algorithm seems to move the prism actuator based on what's happening in the scene. At least for some cameras I've tested.

Did IQ's suddenly drop sharply while I was away?
I enjoy 3D in spite of HDMI 1.4!
Full screen only 3D doesn't cut it!
Frank is offline  
post #635 of 1097 Old 05-16-2012, 07:30 AM
AVS Club Gold
 
Don Landis's Avatar
 
Join Date: Jun 1999
Location: Jacksonville, FL
Posts: 10,761
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 37 Post(s)
Liked: 106
Wolfgang- Frank is spot on about OIS. I have always disabled it when going to the tripod with two cams. I have even noticed OIS lag issues when panning a single cam in 3D. Have to look close for it but it's there. I kind of wish I could enable the OIS with a button but can't have everything.

The camera alignments through zoom ranges is indeed an issue. As a matter of practice, I rely on the Vegas auto correct key frames to make the clip work through a zoom range if I intend to use it. That usage happens only about 10% of the time as I normally zoom to a new focal length just for a static shot.

Quote:


Yes, up to now I am going for parallel axis. But there are limitations that I have become aware with the first tests - if the IO becomes to large, Vegas is not able any more to adjust the parallel axis to move everything behind the stereo window. So I think that I will tend to go more for far point parallax, especially with larger IOs.

But this is where I resort to my stereo base rules and use my calculator.I recall Frank doesn't like "rules" and would rather wing it. But secretly I believe he just knows what works and what doesn't. These limits probably match what I get from the calculator anyway.
You are correct in the observation that Vegas can't move the scene if the stereo base is too large. But when that happens you probably are outside the range of the stereo base for the stage size you selected. I have a couple shots in Valley of fire like that. Valley of fire has lots of close up stages that I shot with too much stereo base and it is upfront well into negative parallax and I can't push it back. When I look back on those shots, I recall shooting with a stereo base of 16 inches ( I slate my shots) but the calculator would have told me the I.A. should have been at 8" With the calculator and a laser range finder I should not make that mistake again. Plus as I shoot more with wide SB I will become as familiar with the ranges as Frank is.
Don Landis is offline  
post #636 of 1097 Old 05-16-2012, 03:57 PM
AVS Special Member
 
Wolfgang S.'s Avatar
 
Join Date: Aug 2011
Location: Vienna/Austria
Posts: 1,131
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 5 Post(s)
Liked: 16
I have tested both to disable OIS, and to use the Manfrotto 323 plates.

My impression with the disabled OIS: yes, it becomes a little bit better, but it does not go away.

And with the Manfrotto 323 plates: they are simpler to handle, their mechanic is not as fine as the ball heads are. But even here the results seems to be similar.

I have tested that with the cross laser only by now. So I have to take some shoots outside if the weather is fine tomorror.

Yes, I agree that the stereobase has been too large very likely. The minimum distance was not taken into account enough. I would be keen to learn more about the stereo base calculator you use.

Kind regards,
Wolfgang
videotreffpunkt.com
Wolfgang S. is offline  
post #637 of 1097 Old 05-16-2012, 04:03 PM
AVS Special Member
 
Wolfgang S.'s Avatar
 
Join Date: Aug 2011
Location: Vienna/Austria
Posts: 1,131
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 5 Post(s)
Liked: 16
Quote:
Originally Posted by Don Landis View Post

I had this nice stereo base calculator for my laptop I had been using but this year have decided to retire the PC in favor of my ipad. But I missed my calculator. As I was waiting for a render to finish up, I decided to research this highly specialized tool in the ipad app store. I found one that looks nice and is easy to use. Anyone doing hyper stereo base shooting will appreciate this:

This looks like it will work on the iphone too.

Don, what is the name of this tool?

Kind regards,
Wolfgang
videotreffpunkt.com
Wolfgang S. is offline  
post #638 of 1097 Old 05-16-2012, 06:03 PM - Thread Starter
AVS Special Member
 
Frank's Avatar
 
Join Date: Apr 1999
Location: Dept. of Offense
Posts: 5,194
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 4 Post(s)
Liked: 14
Quote:
Originally Posted by Wolfgang S. View Post

Don, what is the name of this tool?

I am wondering the same thing.
Perhaps this is it?

It's free for IPAD.

Did IQ's suddenly drop sharply while I was away?
I enjoy 3D in spite of HDMI 1.4!
Full screen only 3D doesn't cut it!
Frank is offline  
post #639 of 1097 Old 05-16-2012, 06:16 PM
AVS Club Gold
 
Don Landis's Avatar
 
Join Date: Jun 1999
Location: Jacksonville, FL
Posts: 10,761
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 37 Post(s)
Liked: 106
Quote:
Originally Posted by Frank View Post

I am wondering the same thing.
Perhaps this is it?

It's free for IPAD.

Yes. There were only two and I looked at both. I preferred this one.



Sent from my ipad3.
Don Landis is offline  
post #640 of 1097 Old 05-17-2012, 03:06 AM
AVS Special Member
 
Wolfgang S.'s Avatar
 
Join Date: Aug 2011
Location: Vienna/Austria
Posts: 1,131
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 5 Post(s)
Liked: 16
Thank you for the hint to this tool - looks nice, and i have installed it on my iphone too.

What I miss a little bit are explainations. The tool seems to work to a similar formula like the Bercovitz formula

http://nzphoto.tripod.com/stereo/3dtake/fbercowitz.htm

but I miss how you bring in convergence. So is the tool designed for parallel axis only?

Kind regards,
Wolfgang
videotreffpunkt.com
Wolfgang S. is offline  
post #641 of 1097 Old 05-17-2012, 11:05 AM
AVS Special Member
 
Wolfgang S.'s Avatar
 
Join Date: Aug 2011
Location: Vienna/Austria
Posts: 1,131
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 5 Post(s)
Liked: 16
I have continued today to test the Manfrotto 323 clams. Well, it is very true to say that they are easier to handle. Using them means to accept variances in vertical disparity, what is still the major weakness of my pair TD10. Regardless if I use the balls or the 323 clams.

Further tests - with disabled OIS - has shown a slight improvement. But testshoots going from full wide-angel into full zoom show still the increase in vertical disparity. It is not strong in first half, but then it seems to increase non-linear. So the good news are that I can use the unit in the wide angel until the light zoom for sure, what is fine since that is the major area for shooting.

So I am still thinking about how to continue in terms fo ballheads or Manfrotto 323 clams. Will see.

Kind regards,
Wolfgang
videotreffpunkt.com
Wolfgang S. is offline  
post #642 of 1097 Old 05-17-2012, 02:44 PM
AVS Special Member
 
Wolfgang S.'s Avatar
 
Join Date: Aug 2011
Location: Vienna/Austria
Posts: 1,131
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 5 Post(s)
Liked: 16
I have finalized my testing - conclusion:


1. the behaviour of the optical system of my pair TD10: well, from wide-angel to full zoom you have 30-31 single steps, the ste-fra-Lanc can address that with a variable zoom speed in a great way. From the first 15-20 steps I see no optical issue, but in the last 10 steps the vertical disparity comes in. That are really good news, because now I know how far I can go into zoom for both adjustment and shooting.


2. the Manfrotto parts are less precise in terms of fine-mechanics, the SIRUI ballheads are better in mechanics. To adjust the ballheads is easier, at least with the method that I have described earlier. To adjust the ballheads for object convergence is easier, due to the plate where you can rotate the system in a very precise way. What should not be done with the ballheads is the manual adjustment of the ball itself - that tends to becomes unprecise. It is better to use a long plate to get the ballheads and the camcorder parallel.

However, as said earlier the Manfrotto 323 parts are much simpler, the weight is lower, and after adjustment they work fine too. So for travelling or trekking I think I will use the Manfrotto 323, where the weight is not an issue I will stick to the ballheads.

Maybe parts like that would be the best solution:
http://www.amazon.de/Benro-BE2018-PC...I2DZY3QMPWJKNL


Question: do you guys tend to shoot with parallel axis or do you converge the axis? Using the RittaiCalc3D is a great help to get low disparity-figures, but doing the convergence correction in Vegas increases the disparity again if I shoot with parallel axis. So I still think about how to hold the disparity low, in the final video.

Kind regards,
Wolfgang
videotreffpunkt.com
Wolfgang S. is offline  
post #643 of 1097 Old 05-17-2012, 03:44 PM
AVS Club Gold
 
Don Landis's Avatar
 
Join Date: Jun 1999
Location: Jacksonville, FL
Posts: 10,761
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 37 Post(s)
Liked: 106
I don't think I am getting the message across.

You need to align the camera wide first for rough adjustment. Then do a final tweak at full zoom. Now when you back out to either full wide or even 50% full zoom the alignment should be quite accurate. But, if you just calibrate full wide and then expect that to remain when zoomed in it probably will lose the calibration you need for that magnification. Again, the point of doing a full wide alignment is to get in the ball park, not that that will be your final tweak. The dovetail quick releases are far more accurate for repeat vertical disparity than the ball clamps. For angle, the repeat accuracy remains until you remove the dovetail on the TD10 bottom.


I always adjust for parallel cameras. Converging on a distant object will screw up the space behind and either need correction for flat or interfering with subject. Besides, for extreme distances, parallel or convergence is a good guess at best anyway.
Don Landis is offline  
post #644 of 1097 Old 05-18-2012, 12:32 AM
AVS Special Member
 
Wolfgang S.'s Avatar
 
Join Date: Aug 2011
Location: Vienna/Austria
Posts: 1,131
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 5 Post(s)
Liked: 16
Quote:
Originally Posted by Don Landis View Post

I don't think I am getting the message across.

You need to align the camera wide first for rough adjustment. Then do a final tweak at full zoom. Now when you back out to either full wide or even 50% full zoom the alignment should be quite accurate.


Don, you got the message accross. But to do the allignment in full zoom did not work for me - due to the optical behaviour of my system. If I do the alignment in full zoom with my system, and then go back to full wide angel and shoot in wide angel, I have significant disparities that I do not like. That was the complication here - because from the theory I am completely with you: I also would make the final adjustment in full zoom and expect that it fits then the wide angel too. For my system that does not work out really.

But what works is to make the fine adjustment with with my system in about 50% zoom with both camcorders - I can drive to such a point with my controller with both camcorders in a simultan way and then I do the alignment.

I think that is a specific behaviour of my system. You are lucky if your two TD10s behave better and you can do the calibration in full zoom, go back to wide angel and then it still fits.


Quote:
Originally Posted by Don Landis View Post

But, if you just calibrate full wide and then expect that to remain when zoomed in it probably will lose the calibration you need for that magnification.


In the end of the day, it is not so much a question of first wide angle or first 50%/100% zoom - but it is a question that the alignment is good enough for a certain range. For you the range is 0-100%, for me the range is maybe about 0-50% only (what is fine since I do not shoot in 100% zoom).



Quote:
Originally Posted by Don Landis View Post

The dovetail quick releases are far more accurate for repeat vertical disparity than the ball clamps. For angle, the repeat accuracy remains until you remove the dovetail on the TD10 bottom.

As a technical guy, I alwas tend to reflect the performance of systems - since I know that the overall performance may be driven by different factors.

The ballheads that I use have a significant better fine mechanical precision then the Manfrotto parts. This SIRUI parts are astonishing in terms of fine mechanics. From that side the ability to reproduce the same settings in terms of vertical, horizontal disparity but also rotation seems to be better then the Manfrotto parts.

That must not mean that the overall results are better. If the constrain in the overall system is the mechanical quality of the optical system, like I see that here, maybe it does not matter and you will end up with similar results.

For the ballheads I had to to learn how to do the adjustment, what I can do and what not. Since that was mixed with the sad behaviour of my camcorders that took me a while. If it is worthwile to do the extra efforts with the ballheads - well, that is something that everybody has to decide for himself.



Quote:
Originally Posted by Don Landis View Post

I always adjust for parallel cameras. Converging on a distant object will screw up the space behind and either need correction for flat or interfering with subject. Besides, for extreme distances, parallel or convergence is a good guess at best anyway.

Right, especially since our systems are not as fine as they would need to be. But it can be done that we do the alignment to parallel cameras, at least with some accuracy. For example, I shoot for a window that is 30-40 m away, in my zoom of 50%. The windows has separete 40 cm halfs - well with a 40cm base it is fine to use that to adjust for the 40 cm. That will not hold for 1000 m for sure but that is something that can be done.

But I see an issue: you shoot with so adjusted parallel cameras, what means that everything is in front of the zero window. So you have to move everything behind the window again in Vegas and adjust the nearpoint to the zero window - fine, but that increases the disparity again. Even if I think that convergenc shooting with convergence to the nearpoint will bring in also a lot of disparity in the farpoint.

The issue becomes larger if the IO you use is huge. So for hyperstereo you will end up with converging cameras, I think. Maybe converging cameras in the farpoint?

Kind regards,
Wolfgang
videotreffpunkt.com
Wolfgang S. is offline  
post #645 of 1097 Old 05-18-2012, 06:37 AM
AVS Club Gold
 
Don Landis's Avatar
 
Join Date: Jun 1999
Location: Jacksonville, FL
Posts: 10,761
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 37 Post(s)
Liked: 106
"But I see an issue: you shoot with so adjusted parallel cameras, what means that everything is in front of the zero window. So you have to move everything behind the window again in Vegas and adjust the nearpoint to the zero window - fine, but that increases the disparity again. Even if I think that convergenc shooting with convergence to the nearpoint will bring in also a lot of disparity in the farpoint."

The negative parallax objects position with respect to zero are controlled by the stereo base and lens focal length. Absent a 3D monitor, I have to rely on the calculator to tell me whether I have to much stereo base or too little. Frank has a clever way to generate a 3D image with his system connecting to a laptop. I have been monitoring the cost of camera mounted 3D auto stereo monitors using hdmi twin input but at NAB the one made by Marshall is still $7500. Besides, this issue of where the convergence point is for long range distant objects is somewhat academic in that how the heck can you determine where to angle your cameras when the distant object is extreme distant. I say for image resolution it is converged. Last night's project shoots the sun. 3D using a stereo base of 26" is hardly going to render a roundness factor to the sun sphere. What is in 3D is the near objects, the water, boats and even clouds in the near by sky. The sun will be flat background. The main danger of convergence is when it happens too close and allows near objects in the shot to cause the viewer to see strange image behavior. So, when I do that full zoom calibrate, I just make sure the object used in the distance does not cross in the viewer of the two cameras.

When I initially calibrated my system using the shims for vertical disparity, camber, and tow of each camera in my shop with a paper target about 15 ft away, I have not had to adjust it since, about 9 months ago. The igus slide tables with Manfrotto quick release keeps consistent calibration. What I do have to adjust is the camera horizontal angle because of removing the little dove tail foot from the TD10. I don't understand why your two TD10's both have so much slop in the zoom range that cause the vertical disparity to go out. That is really strange how that could happen. Frank, do you have any ideas? I can see focus issues due to a bad back focus setting, a factory calibration but not a change in vertical disparity. It's almost as though the lens moves up and down when you zoom. Wait a minute, I think I know what is wrong. If your cameras aren't on the same level plane, they could be adjusted using your ball heads to set vertical on a distant object. This will indeed cause a change in vertical disparity through a zoom range. It could be your cameras aren't on the same level plane. Since the TD10's are more like a boat and don't have any flat level surfaces with respect to the lens center line of view, you have to rely on optics to adjust the level. If you adjust one camera to a target center point at one distance (full zoom for accuracy) and then move the target distance maintaining height, the center point should not move. If it does, then your camera level plane is off. You can get away with this bad alignment in your calibration because the Vegas auto correct fixes this in post. But, as you said the slop in the zoom alignment of your system is too much for your comfort. Consequently, you choose to limit the zoom range in your production. I chose to spend quite a bit of time calibrating these issues once, and then I have the full range. I know Frank did too because he once told me his middle name is "precision".

The real question now is do we need all this precision? Probably not but as in most cases, it's a question of compromise. My NEX5n shoots 3D by stitching together hand held images as I sweep the camera in an arc. Not precise at all but the results look fine. You have to compromise zoom range, I compromised initial shop time to align the design for precision and don't permit any field adjustments, but I don't need them. The NEX 5n system compromises pixel size as it crops the pictures for the stitch. If you used the full zoom range without realigning everything with the ball heads you would be needing to crop more to get results too. I think I have a better understanding now where all this ends up.

Sent from my iPad3 so please excuse the spelling errors.
Don Landis is offline  
post #646 of 1097 Old 05-18-2012, 07:27 AM - Thread Starter
AVS Special Member
 
Frank's Avatar
 
Join Date: Apr 1999
Location: Dept. of Offense
Posts: 5,194
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 4 Post(s)
Liked: 14
Here is my best effort at a SWAG of what's causing the variable vertical disparity throughout the zoom range.....
It is impossible to manufacture two camcorders that are exactly optically identical.
There will always be a variation between the optical path center-line and the perpendicularity to the CMOS imaging chip.
Any deviation from perfect perpendicularity will cause a change in the vertical alignment over the entire zoom range.
I suppose you could compensate by placing one camera higher then the other by the exact same proportion as the internal deviation between the camcorders.
However this would not be practical IMO.

Did IQ's suddenly drop sharply while I was away?
I enjoy 3D in spite of HDMI 1.4!
Full screen only 3D doesn't cut it!
Frank is offline  
post #647 of 1097 Old 05-18-2012, 09:57 AM
AVS Special Member
 
Wolfgang S.'s Avatar
 
Join Date: Aug 2011
Location: Vienna/Austria
Posts: 1,131
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 5 Post(s)
Liked: 16
Well Frank, I tried to locate one camcorder higher - also a German friend in the German Stereoforum had had that interpretation. But up to now I failed here. The issue could be that it is hard to estimate the difference in the heigh. And then I am not so sure if that is really the issue, since the behaviour seems to be non-linear. But maybe I will test that again - I have ordered some additional, longer 1/4 zoll screws to be able to play with that.

Don, yes it is a pitty that the clever solution from Frank will not work for you and me. The issue is that Sony has disabled the composite output for the TD10 if we use AV/R. Older Sonys, or even the professional version of the TD10 does not have that limitation. I would also like to have a box that gives me both the preview in 3D, and allows the capturing. If you like a Mac you could do that with Blackmagics Ultrastudio 3D

http://www.blackmagic-design.com/pro...ultrastudio3d/

since it uses a thunderbold connection.


Quote:
Originally Posted by Don Landis View Post

When I initially calibrated my system using the shims for vertical disparity, camber, and tow of each camera in my shop with a paper target about 15 ft away, I have not had to adjust it since, about 9 months ago. The igus slide tables with Manfrotto quick release keeps consistent calibration.

Yes, I believe that. I have done that only yesterday, did the calibration with a right angle valve - and that is quite stable with the Igus tables.

Quote:
Originally Posted by Don Landis View Post

What I do have to adjust is the camera horizontal angle because of removing the little dove tail foot from the TD10.

That is some advantage with the ballhead - since the foot parts are quite stable and result in the same angel - that can be adjusted with a separate control knob. The scale is not fine enough for our purposes - only 5 Degree scale. But you can adjust it to an object - theoretically for convergence or to adjust to the IO.

Yes I like the iphone calculator - that is a nice tool.

Quote:
Originally Posted by Don Landis View Post

I don't understand why your two TD10's both have so much slop in the zoom range that cause the vertical disparity to go out. That is really strange how that could happen. Frank, do you have any ideas? I can see focus issues due to a bad back focus setting, a factory calibration but not a change in vertical disparity. It's almost as though the lens moves up and down when you zoom.

I think Frank is right - that is an production issue. Do not forget that the way how we use the camcorders is not the way for that they were designed.

Quote:
Originally Posted by Don Landis View Post

Wait a minute, I think I know what is wrong. If your cameras aren't on the same level plane, they could be adjusted using your ball heads to set vertical on a distant object.

If they are not on the same level plane, the ballheads are able to adjust that by looking up or looking down a little bit. But that will be valid for one distance point only. If you zoom in or zoom out, you will loose this calibration again. So my idea was to bring up one of the camcorder to a higher level plane by adding small metall slices (damn, I am not sure about my translation here). But such slices are too thick maybe. So I tried some aluminium foil, but that is hard to do if you do not know how much is missing.

And what confuses me still is that fact that the behaviour seems to be non-linear. If the issue is that he optical axis are not at the same level plane, it should be linear I think.

Quote:
Originally Posted by Don Landis View Post

If you adjust one camera to a target center point at one distance (full zoom for accuracy) and then move the target distance maintaining height, the center point should not move. If it does, then your camera level plane is off.

To move the target distance - hard to do that so precise that you get a clear result. What you can do is to use the zoom out to check that. I think I will give that a try.

Quote:
Originally Posted by Don Landis View Post

You can get away with this bad alignment in your calibration because the Vegas auto correct fixes this in post. But, as you said the slop in the zoom alignment of your system is too much for your comfort. Consequently, you choose to limit the zoom range in your production. I chose to spend quite a bit of time calibrating these issues once, and then I have the full range. I know Frank did too because he once told me his middle name is "precision".

Well, I am a technical engineer, and I have not given up yet. And it is always fun to play with such systems.


Quote:
Originally Posted by Don Landis View Post

The real question now is do we need all this precision? Probably not but as in most cases, it's a question of compromise.
Sent from my iPad3 so please excuse the spelling errors.

Well, I have spend now a lot of time and use the cross laser and a white wall for my geometrical exercises. That takes time that I would spend for editing or shooting. But it still is fun.

Kind regards,
Wolfgang
videotreffpunkt.com
Wolfgang S. is offline  
post #648 of 1097 Old 05-18-2012, 10:13 AM - Thread Starter
AVS Special Member
 
Frank's Avatar
 
Join Date: Apr 1999
Location: Dept. of Offense
Posts: 5,194
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 4 Post(s)
Liked: 14
Quote:
Originally Posted by Wolfgang S. View Post


And what confuses me still is that fact that the behaviour seems to be non-linear. If the issue is that he optical axis are not at the same level plane, it should be linear I think.

Actually, I believe it is a LOG function and what it's doing makes perfect sense to me.

Did IQ's suddenly drop sharply while I was away?
I enjoy 3D in spite of HDMI 1.4!
Full screen only 3D doesn't cut it!
Frank is offline  
post #649 of 1097 Old 05-18-2012, 01:47 PM
AVS Special Member
 
Wolfgang S.'s Avatar
 
Join Date: Aug 2011
Location: Vienna/Austria
Posts: 1,131
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 5 Post(s)
Liked: 16
Unfortunately, it makes not really a sense. The reason why I believe that is that one of the TD10s behave linear, but the other does not. I took me 3 hours to measure the optical system of both cameras, and I can confirm significant different behaviour of the two optical systems (OIS was off during measurement).

So the simple answer is: both camcorders work fine as a single unit. But to pair them is not the best idea - even if I adjust them in wide angel. Have to think about how to overcome that. Looking back it would have been a good idea to do a classical pairing activty.

Kind regards,
Wolfgang
videotreffpunkt.com
Wolfgang S. is offline  
post #650 of 1097 Old 05-18-2012, 03:26 PM
AVS Club Gold
 
Don Landis's Avatar
 
Join Date: Jun 1999
Location: Jacksonville, FL
Posts: 10,761
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 37 Post(s)
Liked: 106
Quote:
Originally Posted by Frank View Post

Actually, I believe it is a LOG function and what it's doing makes perfect sense to me.

You and I think alike! I did some quick and dirty sketching of the height disparity as a function of the zoom ratio picture size area and it is not a linear change.

But, Frank, I do believe that while you are correct that it is impossible for optical precision to ever be at zero tolerance, we are comparing the precision of production optical blocks in the camcorder vs. the precision of setting two ball heads for precision plumb setting to keep the optical centerline level. I put my bets on the optical block construction precision being far better than the ball head being plumb.


Wolfgang: translation of slices or what I call shims- I think we are talking the same thing. I sent you a photo with how the shims are used. Depending on the vertical and tilt of the camcorder, adding a very thin strip of metal between the two flat surfaces, locking the quick release base tight and then checking calibration is what I did. I kept making changes to the thickness if each shim and checking the cross hairs on the wall target for each camera to get the level perfect at full telephoto. Then pulling out wide, it just gets smaller on the screen but doesn't change position. This took me a whole afternoon but when I was satisfied, it was a permanent construction.
The mounting of the foot is eyeballed to align it so it puts that camcorder perpendicular to the bench. I do my best in the field if I need to attach the dove tail piece to the camcorder but making it snug and then just doing a little forced twist to line up on a far distant object is all that it takes. The vertical and tilt does not vary. I would think that if you loosen a ball claw to make your field adjustment you would through out every directional calibration. A more difficult setting to complete. In a practical sense example, I arrived at my beach front shooting location and had my hyper stereo base set up and aligned to a sea marker maybe a mile off shore at full telephoto in under 3 minutes. This was everything including pulling the camcorders out of the case and setting up the tripod, bench, sliding the igus tables on and connecting the Lanc shepherd. I was ready to go. In the setup, I began to set the sea marker to a left upper cross hair at full zoom on the left camera. Then I checked it on the right camera and it was dead on vertically, but the angle was off by about 3 mm on the Sony monitor. I twisted the right camera so the cross hairs was just about the same distance putting the sea maker to the left side of the cross hairs. The sea marker was 1 mile away. It's width is about the same as my bench stereo base, a convenient target.
Don Landis is offline  
post #651 of 1097 Old 05-19-2012, 04:32 AM
AVS Special Member
 
Wolfgang S.'s Avatar
 
Join Date: Aug 2011
Location: Vienna/Austria
Posts: 1,131
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 5 Post(s)
Liked: 16
Quote:
Originally Posted by Don Landis View Post

I put my bets on the optical block construction precision being far better than the ball head being plumb.

Then you have lost this bet! Yesterday evening I spend 3 hours to measure the optical systems of both TD10s, and today additional 3 or 4 hours. I did that by plotting the position of the crosslines of both camcorders to a large piece of paper on a wall, using the cross-laser to transfer the position of the cross hair to paper. I did that for both camcorders. And did that for different zoom positions and used the rig with the ballheads, and saw again that this is precise really.

What was transfered to paper is the position of the 4 corner points of the rectangle of the cross hairs. Then you can follow how the position of the corner points changes if you change the zoom. A variance in the corner points will end up in a variance in the final video, depending how you adjust the cameras. The variance can be both vertical and horizontal. In other words: the result shows the differences of the two optical systems when you compare the two camcorders.

I did the test with two strategies for the start point:
1. I adjusted the ballheads in a perfect way, installed the camcorder and had a look to the level of adjustment both in full zoom and in full wide angel- and for a lot of positions between. This gives me an information about the quality of the mechanical system but also about the optical system.

2. I did the same as in 1., but then adjusted the camcorders in full wide angel and measured again the adjustmet both in full zoom until full wide angel - and for a lot of positions between. This test gives me an information about the behaviour of the optical system.

Overall results:
a. The mechanical system is really precise, if you adjust the tripod and the igus system is in balance. Its behaviour is reproducible.

b. There is a variance both in horizontal but also in vertical alignment in the optical system. If you start with an adjusted system in full wide-angel, you find a non-adjusted system in full zoom and the other way round.

c. The TD10 has 30 zoom steps. With an adjusted system in full wide angel as in 2., you can zoom in and stay adjusted for the first 5 steps in a perfect way, until 10 in an acceptable way - and then you see that you shift out of the adjustment more and more.

For the test 2 I have calculated the pixel deviation - from the 1920 pixel I see a variance of 84 pixel or 4,4%, from the 1080 pixel I see a variance of
18,8% or 203 pixel (so, measured if I adjust the both cameras in full wide angel and then zoom full in. This was measured with the displays, an overscan is not considered here). Maybe only valid for the specific setting of my measurement - but just to have a feeling.

Next I have tried to calculate the difference in the heigth, if that causes the troubles. Using the crosslaser again to measure the difference in full zoom, and using the theorem of intersecting lines (is that translation ok - dont know) I end up with a height difference of 4.1 mm that I would have to bring up the lower camcorder. I am not sure if that is right or not, the figure looks high to me - but it is worth a try. I wait for longer 1/4 inch screws, they are ordered - and I have small metall plates with that 4.1 mm here.

Overall conclusion: if I continue to use this camcorder system, I have to do the adjustment in full wide angel or at step 5 to 10 of the zoom. And can use the system from wide angle to the first 10 zoom steps. Or I exchange the camcorder system if the increase in heigth does not work out.

Kind regards,
Wolfgang
videotreffpunkt.com
Wolfgang S. is offline  
post #652 of 1097 Old 05-19-2012, 05:20 AM
AVS Club Gold
 
Don Landis's Avatar
 
Join Date: Jun 1999
Location: Jacksonville, FL
Posts: 10,761
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 37 Post(s)
Liked: 106
I am more convinced you don't understand the process. The cameras do not have a center line cross hairs. Unfortunate because you have to use the off center cross hairs as a relative reference only. Use any one of 4 cross points as a reference between the two, not the accuracy of the zoom within a single camcorder. This is not uncommon that a process is so simple one will overlook it and be off on a mission to blame something like the camcorders not being accurate. If you had a reference to a CENTER SCREEN CROSS HAIRS life would be easier and you would see the camcorder maintains center as you zoom. but as the 4 cross hairs are not centered the camcorder zoom expands and contracts ABOUT THE CENTER. Therefore you can only use the cross point as a reference between the two when both cameras are at either full wide or full zoom but not somewhere in between unless you know the sync device is maintaining zoom sync. My Lanc Shepherd does this nicely.

It too bad you spent so much time trying to prove the camcorders are defective, as the time would have been better spent aligning the quick releases once and done, permanent as accurate as both Frank and I have achieved. I thank Frank for convincing me these were the way to go. I like simple, quick and easy and these are much easier than ball heads and set screws as I tried both before going the quick release way.

If you have ever aligned a scope on a rifle with a laser bore sight the process is much the same, using the bore sight as one camera and the scope as the other.

Frank has a video that demonstrates that the quick releases are accurate and repeatable, but maybe I need to make a video on how to build and align the system we use. Im convinced we are losing something in the language translation. There is no reason to have to do any long calibration in the field to compensate for zoom.
Don Landis is offline  
post #653 of 1097 Old 05-19-2012, 06:29 AM
AVS Special Member
 
Wolfgang S.'s Avatar
 
Join Date: Aug 2011
Location: Vienna/Austria
Posts: 1,131
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 5 Post(s)
Liked: 16
Don, I have never talked about one center cross hair, I know that such a thing does not exist. What I do is to track the position of the four corners - so I have have found a way around that in my measurement.

And for sure the issue are not ballheads or quick release plates - I have tried both.

Here is also a picture of my tests. You see the two small rectangles, what are the positions of the cross haircut - for each of the two camcorders in full zoom. The large rectangle is the position of the rectangles of the full wide range for both camcorders. Here the camcorders are aligned in full wide range - so that is nearly the same.

And the lines between show how the positions of the rectangles move when you zoom out to the wide-angel. And yes, it is possible to measure the position of the points between the full wide-angel and the full-zoom.

I full zoom the different position of the two small rectangles show the mismatch in the allignment.

Hope that makes it better understandable.


Kind regards,
Wolfgang
videotreffpunkt.com
Wolfgang S. is offline  
post #654 of 1097 Old 05-19-2012, 06:42 AM - Thread Starter
AVS Special Member
 
Frank's Avatar
 
Join Date: Apr 1999
Location: Dept. of Offense
Posts: 5,194
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 4 Post(s)
Liked: 14
Wolfgang,
Would it be possible for you to graph the x and y deviation as a function of zoom factor for each camera?
Thanks

Did IQ's suddenly drop sharply while I was away?
I enjoy 3D in spite of HDMI 1.4!
Full screen only 3D doesn't cut it!
Frank is offline  
post #655 of 1097 Old 05-19-2012, 08:02 AM
AVS Special Member
 
Wolfgang S.'s Avatar
 
Join Date: Aug 2011
Location: Vienna/Austria
Posts: 1,131
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 5 Post(s)
Liked: 16
Frank, that is the the graph included. You see the numbers 5, 10, 15, 20 and 25. It starts with zero (a number that is missing) and ends with 30/31 (what is in the center of the graph). Sorry, I forgot to mention that.

So, you see that at 0 or 5 you are near a deviation of zero. At 10 you see a small deviation, both in x and y (it is simple the distance between the curves, one time in the x- and one time in y-direction). At 30/31 the deviation has become the 4 mm in x (of the 30 mm of the rectangle) and the 9 mm in the y direction (of the 17 mm of the rectangle), that I mentioned earlier.

Kind regards,
Wolfgang
videotreffpunkt.com
Wolfgang S. is offline  
post #656 of 1097 Old 05-19-2012, 08:09 AM - Thread Starter
AVS Special Member
 
Frank's Avatar
 
Join Date: Apr 1999
Location: Dept. of Offense
Posts: 5,194
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 4 Post(s)
Liked: 14
Quote:
Originally Posted by Wolfgang S. View Post

Frank, that is the the graph included. You see the numbers 5, 10, 15, 20 and 25. It starts with zero (a number that is missing) and ends with 30/31 (what is in the center of the graph). Sorry, I forgot to mention that.

So, you see that at 0 or 5 you are near a deviation of zero. At 10 you see a small deviation, both in x and y (it is simple the distance between the curves, one time in the x- and one time in y-direction). At 30/31 the deviation has become the 4 mm in x (of the 30 mm of the rectangle) and the 9 mm in the y direction (of the 17 mm of the rectangle), that I mentioned earlier.

Thanks.
Unfortunately the image lacks sufficient detail for me to read it with my eyesight.
What about the other camcorder?

Did IQ's suddenly drop sharply while I was away?
I enjoy 3D in spite of HDMI 1.4!
Full screen only 3D doesn't cut it!
Frank is offline  
post #657 of 1097 Old 05-19-2012, 11:12 AM
AVS Special Member
 
Wolfgang S.'s Avatar
 
Join Date: Aug 2011
Location: Vienna/Austria
Posts: 1,131
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 5 Post(s)
Liked: 16
Your are right, the image is of poor Quality - it was a snapshoot between a lot of work that I had to do after lunch.

Maybe it is really a good idea to calculate the deviation, but that should be expressed in pixel I think. Will take a little bit time.

The image shows both camcorders, Frank. You see two rectangles with different positions in the x-y-chart. The deviation is the difference between the position of the rectangles or the lines.

Kind regards,
Wolfgang
videotreffpunkt.com
Wolfgang S. is offline  
post #658 of 1097 Old 05-20-2012, 10:13 AM
AVS Club Gold
 
Don Landis's Avatar
 
Join Date: Jun 1999
Location: Jacksonville, FL
Posts: 10,761
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 37 Post(s)
Liked: 106
I can see the target fine you nicely documented. An excellent piece of work if it had any meaning to resolve the issue you are experiencing. Unfortunately, it appears more to be a precise measurement of a procedure that I feel doesn't work. A procedure that's difficult and and time consuming to align in the field and uses hardware that is bigger and heavier than necessary. If your ball heads could demonstrate a better way to get calibration and more precise settings for disparity between the two cameras, that once set in the shop never need calibrated again, I would be game to reorder the ball heads and give them another try. But I don't see any advantage. Regardless how impressive your charting efforts are, it doesn't resolve the primary goal to make a system that is light weight, portable, accurate, and repeatability for field setup.

If you could, I would like to see your set up ( with ball heads) on a tripod ready to shoot. Maybe what I imagine your design to look like is not what you've done.
Don Landis is offline  
post #659 of 1097 Old 05-20-2012, 10:14 AM
AVS Special Member
 
Wolfgang S.'s Avatar
 
Join Date: Aug 2011
Location: Vienna/Austria
Posts: 1,131
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 5 Post(s)
Liked: 16
Quote:
Originally Posted by Don Landis View Post

The cameras do not have a center line cross hairs.

Dons comment about the missing center line cross hair brought another idea to my mind. I put a simple cross on a piece of paper on the wall and checked the behaviour if you zoom in and out. Well, the camcorder has to be very precise adjusted to the cross on the wall. But that can be done.

Theoretically, the position of the cross should stay at a constant position.

The result was quite clear: the older TD10 keeps the position stable as to zoom position 8-10. Then the cross shifts away. The new TD10 keeps the position much longer quite stable - nearly to the fool zoom position.

Tests to compensate that by increasing the heigth of the lower camcorder has not been successfull. For our purpose we need two camcorder with a good behaviour in terms of optical axis. So the decision is clear: I will change the older TD10 against a new one.

Kind regards,
Wolfgang
videotreffpunkt.com
Wolfgang S. is offline  
post #660 of 1097 Old 05-21-2012, 06:33 AM - Thread Starter
AVS Special Member
 
Frank's Avatar
 
Join Date: Apr 1999
Location: Dept. of Offense
Posts: 5,194
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 4 Post(s)
Liked: 14
Quote:
Originally Posted by Wolfgang S. View Post

Dons comment about the missing center line cross hair brought another idea to my mind. I put a simple cross on a piece of paper on the wall and checked the behaviour if you zoom in and out. Well, the camcorder has to be very precise adjusted to the cross on the wall. But that can be done.

Theoretically, the position of the cross should stay at a constant position.

The result was quite clear: the older TD10 keeps the position stable as to zoom position 8-10. Then the cross shifts away. The new TD10 keeps the position much longer quite stable - nearly to the fool zoom position.

Tests to compensate that by increasing the heigth of the lower camcorder has not been successfull. For our purpose we need two camcorder with a good behaviour in terms of optical axis. So the decision is clear: I will change the older TD10 against a new one.

The fact that Don sees the graph with no problem clearly demonstrates how bad my vision is.
I decided to postpone my retina surgeries until my Medicare coverage kicks in in a couple of months.
Over the last 3 years I have experimented with a lot of combinations of dual camcorder rigs involving over 15 camcorders.
They all had one thing in common which is an optical deviation throughout the zoom range.
I suspect the best chance of getting two that are extremely similar is to make sure their serial numbers are consecutive or as close as possible to each other.

Did IQ's suddenly drop sharply while I was away?
I enjoy 3D in spite of HDMI 1.4!
Full screen only 3D doesn't cut it!
Frank is offline  
Reply 3D Source Components

User Tag List

Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page


Forum Jump: 

Posting Rules  
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off