When small interaxials just don't cut it! - Page 27

Thanks for confirming that Frank--

Then that begs the next question for the later parts where there appears no miss alignment between the two cameras, where's the problem? The falling peanut as just one example is not unique to your video, but appears to be the norm. The only issue I have with shooting this subject is that the standard video of 29.97 fps is just too low a frame rate for the action. But even genlocked cameras would have this issue.
All I have to say is congratulations, your video for 29.97 fps appears perfect!

Here is a webinar on 3D from Tektronics

The webinar begins at the very basics of 3D. Most of us are well beyond this but continue and you will find quite a nice advanced level discussion on 3D with some tips to avoid trouble in 3D production.

Last I checked my PS3 did nothing more than advance one frame at a time at various rates. So, it appears to me just like advancing on the timeline one frame at a time. The method in Vegas to slomo has some superiority to that by creating "in between" frames from the original frames as key frames. This can smooth out the slomo considerably, although the between frames are artificial. If you haven't done this yet, one method in Vegas is to take a short clip you slice out, say 5 seconds long and stretch it to say 15 seconds long, a 1/3 slomo. This will generate two additional between frames and give you a smoother look. You stretch the clip by holding down the control key and grabbing the right edge of the clip and dragging it out. As you do that there will be a sawtoothed line in the center to denote the stretch. Of course this is just the proxy on the timeline so you have to render the output to see the final results. This process does not modify your original files.

Yes, Know the drill with that. Haven't looked at your file specifically, but used the PS3 tick tick tick slomo before. Many of the DVD players work the same.


Hey- I did find some links for you to dream on-

http://www.visionresearch.com/Produc...Speed-Cameras/ Look at the flex camera specs. I saw 2 of these in a SBS rig on display at NAB.

Here is a v-blog from the 3DGuy-

http://www.youtube.com/watch?feature...&v=sb-dfX5hv7c

I thought the Volt was for her?

Really? Mine has not detected that wording yet!

Recall a year ago, Frank said his budget for cameras was unlimited. I wonder what changed. Wife got her Volt?

Have you seen such a box? I have seen twin SDI-Inputs, or one hdmi and one SDI input? But I have not seen 2x hdmi input and 1x hdmi output.

I see Marshall has added a note about cameras requiring genlock to their spec page after our discussion at NAB.

Blackmagics offers at least something for the field:

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

and then that box:

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

But I do not like that they run via thunderbold only.

From the specs I saw, the USB is 2.0 and for software updates. Are we talking the same thing?

Two of these would get you the sdi needed to feed the Marshall auto stereo monitor, but the cameras still need to be genlocked, or use a 2x hdmi to 2x hdmi frame store. What Frank needs is a frame store with hdmi in and out plus a 3D combiner for frame packing or SBS. I haven't seen anything like that.

I think Frank will ultimately break down and shoot with that Flex twin camera rig plus the auto stereo monitor from Marshall. It will do everything he needs. High speed, genlock, and variable IA with field 3D autostereo monitor and scopes.

I think I would maybe be the only customer. Meanwhile, my plan is to extend the sync cable and build a rig for distance shots with the appropriate interaxial distance for the highest level of magnification available in the aftermarket lenses. I'll use an external monitor to frame the shots and crop them a little if necessary in post. I'm using my TD 10 for everything that is within its zoom range. As you can see, I am fairly limited in technology, but am working to leverage what I do have.

(I have gotten a few good aerial shots with my HD Heros mounted on my octocopter.)

Blackmagic Ultrastudio 3D does not have an usb 2.0 connection but only a thunderbolt connection.

http://www.blackmagic-design.com/pro...io3d/features/

That could be necessary since it is the only Blackmagic solution that offers to handle two 3D streams. I do not know exactly what you mean by frame store - but I know for sure that the Ultrastudio 3D has 1 x HDMI type A connector including 3D frame packing support. There you have a 3D playback function.

What I do not know is if it allows a live preview in 3D too or if it is a playback funktion only.

When thunderbolt will be available for 3D Laptops, this could become an interesting alternative to Marshall. Because in addition it offers capturing of uncompressed or low compressed material... But at the moment, there is no Windows driver, it works only for Linux and Mac.

Wolfgang- OK, I failed to look at the Ultrastudio3D specs. I was referencing the HDMI to SDI converter box for that USB 2.0 comment.

The Ultrastudio3D seems to be the box for the flex cam system but not for consumer camcorders, I'm afraid. The tech specs indicate this does not contain a frame store but just a sync lock, so your sources and this device all need common sync. I also wonder if this will work with twin cameras and combine them for stereo 3D. Need to read more.


Its a pretty common device used in broadcasting and became very popular in the late 80's for small production houses to have video mixers such as the Video Toaster. A Time base corrector will stabilize a video tape player to another Video tape player for creating dissolves. These tape players must support advanced sync to use a TBC. Absent this, such as most non-broadcast devices a frame store synchronizer is used which will work for asynchronous video sources absent sync locking capability such as consumer and most prosumer camcorders. Originally the frame store synchronizer was used in TV stations to combine external video with house video. The ultrastudio3D has an input for blackburst and trisync telling me it is a genlockable device absent the frame store circuitry. Therefore it would not work with a twin consumer camera system for 3D. The device accepts sync in on pin 6 of the breakout cable connection as well as analog video connections for cameras that do not support SDI or HDMI.

This is close but doesn't appear to be quite ready for what we are needing yet for consumer camcorders.

i used an LCD passive 3D TV to ensure there was no difference in synch phase attributable to watching with shutter glasses.

I find it remarkable that Don and his wife were not distracted by the lagging of the Right image at the beginning of Frank's Blue Jay video (bluejay fights.mt2s). For me it was almost unwatchable! It was a jumbled mess for my eyes. This is an illustration of how human perception of stereoscopic 3D varies widely.

The correctly aligned version (resync.m2ts) I found quite ok to watch, though I was conscious that the frame rate could have been higher to capture the "quick snappy movements", as Don has mentioned. [Or perhaps 60p instead of 60i?]


Frank,
yes i for one would be interested if you did have a real world example. Given that a one frame difference is almost unwatchable for me, i wonder what say a 1/10th frame difference would look like as video, with a fast moving subject.

Jeff

Jeff-

I'm pretty sure I know why the video here looked acceptable to our eyes. Consider that the scene action by the animals was only out of sync for 1 frame and then the animals held still for quite some time. This allowed the two images to resync naturally. It was the strobe like movement that allowed us to watch in a natural unaffected way.

By contrast, had the two images been out of sync for every frame I can assure you both of us would have wondered why the animals were so "blurry"
There was another thread in this forum ( not the 3D components section) that was discussing a single frame that was doubled once every couple of minutes. As I recall, two posters made such a huge brouhaha over it posting hundreds of times that the single extra frame inserted every couple of minutes made then sick to watch. I don't know if one was feeding off the other but it's an example how, as you stated, something that can go unnoticed by some people are completely nauseating to others.

But aside from all that, Frank's blue jays and squirrels were quite entertaining to watch, but it clearly demonstrated the need for high speed video to catch the real motion that happens much too fast for 29.97 fps. 60p would have helped but my guess is we need more like 500 fps or more to really catch the snappy movements.

Don,
I do not demand perfection. For me a repeated frame every 2 minutes (if noticed) would be an annoyance, but I could enjoy the other frames.

The effect I see in the first 3000 frames (1m 40sec) of bluejay fights.m2ts is tolerable for the background leaves gently fluttering in the wind. As for the quickly moving wildlife, with every burst of movement I perceive a watery mirage-like effect, such as you might see in real life in the air above a hot road.

This effect is quite disconcerting and disturbing for my vision. A trial to watch. As I say, that this part of the video did not particularly worry you and your wife, really does show how human 3D perception varies.

Yes the blue jays are very entertaining to watch. I'd be interested in 3D captured at 120fps, though the issue arises of a suitable display/projector!

Don,

I do not know it the Ultrastudio 3D would be userable for our twin-consumer cameras really. But I understand that it should be possible to capture the footage from hdmi with one unit - to two streams.

The charme could be to use that as a preview unit, but here I am not sure if it would work really. To capture to different low-compressed streams should be possible too - but here one could use other units too. It should work from the cameras, they offer an uncompressed footage from the hdmi interface (even if I can hope only that it did bypass the hardware-compression).

Thank you for the explaination about the frame store unit - but if I understand you right that does not exist for our actual systems?


Jeff,
for me the first video with a missync of 1 frame was unwatchable too. Of the 3D effect was reduced due to the irritation that was caused by that. I have no clue about how many ms missync are fine really, for such a fast moving object. Would be nice to understand and know that better.

Highspeed cameras for 3D would be nice, but I think that is very expensive. And then again the question how to pair that!

Jeff- if I didn't step through the video frame at a time looking at the side by side I would not have detected the blu jay one frame behind every so often as it made its quick move and then remained still. The continuous movement of the leaves one frame out does give a blur as you said, the look of the watery flow. But, it was also out of focus because of the narrow depth of field.

Wolfgang- I'm not aware of any frame store devices for hdmi. Harris makes them for component analog and sdi. So for consumer camcorders, I know of none.

Cost is an individual issue, Many people think what I spend on video is expensive but I see what Frank spends and that is out of my league. But, that is about one's hobby. As a business, I spent 6 figures for equipment but the clients really paid for all of it. So, buying a pair of flex Phantom high speed cameras and a 3D Ultra Black Magic box to record and the Marshall 3D field monitor seems all relative to your market, certainly not in my budget as a hobby. And, that would be a system to pair everything and have the results recorded with live monitoring. It exists and I've seen the system. Besides, my projects don't really require such a system. Frank has demonstrated his do! He just has to decide if it is in his budget.

I also think the system exists to do high speed 3D with consumer level budget but the compromise would be the need to correct stuff in post and this was a compromise that Frank said he couldn't live with. So, with today's technology, he is in the broadcast camera and system camp.

There is no reason why everyone should see stereoscopic 3D in the same way. The background of the video did not particularly trouble me. The movement of the birds and squirrels did - enormously!

At the 1m40s point, with a cut to a different point in time, I could immediately see an improvement, and relax. [The first time through I didn't even reach the 1m40s point; I felt so uncomfortable I stopped watching.]

I think this is a very significant finding: a timing discrepancy at the start of that video which for Wolfgang and me was practically unwatchable, but which was not distracting for you and your wife.

_________

CREATING A TEST FILE

This may be an unnecessarily complicated approach but it is an experiment I did last year.


Here is an AviSynth script I loaded into VirtualDub:

V=DirectShowSource("Filmvideoat23.976fps.mkv", convertfps=true)
copyv=DirectShowSource("v2Filmvideoat23.976fps.mkv", convertfps=true)
Expansionfactor=1 # Use if source is full width side by side
Expansionfactor=2 # Use if source is half-width side by side
Quality=2 # For full width source at full quality, or to retain vert resolution with half-width source
Quality=1 # Normal quality setting, for much faster processing
w=width(V)
Halfwidth=w/2
h=height(V)
# In next line, set whole number percentage of image width you wish to retain (minimum 50)
Retain=60
crophoriz=(100-retain)*w/400
Newwidth=Expansionfactor*Quality*(w -4*crophoriz)/2
Newheight=Quality *h/2
v1=selectevery(v,2,0,0,1,1) # selects frames 0, 0, 1, 1 of file v1 and then increments by 2
v2=selectevery(copyv,2,0,1,1,2) #selects frames 0, 1, 1, 2 of file copyv and then increments by 2
Left=V1.crop(crophoriz,0,-(halfwidth+crophoriz),0)
Right=v2.crop(halfwidth+crophoriz,0,-crophoriz-2,0).addborders(0,0,2,0,color_darkslateGray)
stackhorizontal(Right,Left)
bilinearresize(newwidth,newheight)
The above script duplicated the frames of a 24fps side-by-side 3D file to create a 48fps side-by-side file but misaligned the Left and Right content by the equivalent of half a frame at 24fps. [The addition of the thin slate gray border was to facilitate cross-eyed viewing.]

The content "==", and its misalignment, can be represented like this:

L=============
R============

I redefined the frame rate as 50fps and saved the file. I viewed it on a 2D TV at 50fps, using cross-eyed viewing. It was of unwatchable quality for my eyes.

Tonight, I loaded up the file and played it on my passive 3D TV. It was so poor as to be of unwatchable quality. The Left - Right phase difference of 1/50th sec, for an underlying effective progressive frame duration of 1/25th second, was too much for my vision to accommodate.

I'll see whether I can come up with a method that introduces a smaller discrepancy. For example, from a high speed side-by side source selecting frames 1, 5, 9, 13, etc for Left; and frames 2, 6, 10, 14, etc, for Right.

Jeff- that experiment is not what happened in the Blujay video since the frame's focused subject matches except during the snap move. This was because there are lengthy moments where the birds are still followed by a frame or two of mismatched motion. Your test example shows every frame mismatched.

When it comes to cross eyed viewing, that really bothers me and I find it too much trouble beyond a brief moment to view. I'll stick to passive and active glasses, thank you. third choice is auto stereo and 4th anaglyph.

The film clip included some panning as well as some static shots where the actors were standing still and talking. The act of talking looked very odd with a 20mS Left-Right discrepancy. Slow panning led to a serious disruption of the integrity of the whole of the frame. I stress this was for my vision. I am looking for a less extreme example of a Left - Right timing discrepancy, to see how noticeable any effect is. It could be that the brain ignores discrepancies below a certain time interval.

Yes, I think I have my brain well trained to ignore certain distractions. Comes from 45 years of marriage, and kids.

It's funny but when I discovered the first clip mismatch, it was not because I was bothered by something, it was because I was looking for what others were bothered by and until my post no one here was specific as to what was the problem. A quick frame by frame revealed the difference. Then the next clip edited into the series was fine and there after. But as I said they first clip didn't bother me or even distract. I was just watching the content.